Talk:Genetic algorithm/Archive 1

This is an archive of past discussions about Genetic algorithm. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page.

Archive 1

Archive 2

Can we add a link to this somewhere?

http://www.wreck.devisland.net/ga/

Absolutelely great example of a genetic algorithm in Actionscript. Didn't add it myself as I'm not exactly sure where to put it :)

--- Your program is great!. find some time to improve it. —Preceding unsigned comment added by 134.102.40.14 (talk) 09:22, 16 May 2009 (UTC)

get some pictures up there

hence —Preceding unsigned comment added by 138.16.57.61 (talk) 20:36, 23 October 2008 (UTC)

Featured Article

I'd like to officially nominate this article for consideration as a featured article.
Tyler 19:12, 17 May 2007 (UTC)

Strong AI question

Doesn't Hubert Dreyfus include genetic algoritms in his 'Strong AI' category, together with neural nets?

Have his 'anti-Strong AI' fellow-travellers John Searle or Roger Penrose ever commented on this?

ericross 15:54, 19 May 2003 (UTC)

Sentence that doesn't make sense

I was reading this sentence and I don't think it makes sense. "Genetic programming algorithms typically require running time that is orders of magnitude greater than that for genetic algorithms, but they may be suitable for problems that are intractable with genetic algorithms."

Shouldn't the two instances of "genetic algorithms" (one immediately before the comma and the last one) be "non-genetic algorithms"?

I have a feeling that I've somehow misunderstood the topic.. Neoncow 20:53, 5 Oct 2004 (UTC)

"Genetic programming" (GP) is different from "genetic algorithms" (GA) due to historical nomenclature. The sentence is supposed be comparing GPs with GAs, but it's very ambiguous as written. I'll rewrite it. --Lexor|Talk 07:03, 6 Oct 2004 (UTC)

Could this bit of jargon be merged here? Building block hypothesis--nixie 00:50, 23 Mar 2005 (UTC)

The building block hypothesis is not accepted by all people in the field, and is best to remain in its own article. A reference to some of the theoretical hypotheses and observations could possibly be put here (there are some getting lost in the text), only having the BB hypothesis would not be NPOV --Anthony Liekens 02:51, 25 August 2005 (UTC)

GAs and Biological Realism

I have a big problem with the following line:

"A problem that seems to be overlooked by GA-algorithms thus far is that the natural evolution maximizes mean fitness rather than the fitness of the individual (the criterion function used in most applications)."

It is the consensus of the biological community that natural evolution acts to maximize individual fitness, even to the deteriment of the group (mean fitness). [for non-professional reading see Richard Dawkins book 'The Selfish Gene']. There has been considerable debate in the past on this issue, and it has been settled in favor of individual fitness. —Preceding unsigned comment added by SteelSoul (talk • contribs) 03:21, 13 October 2007 (UTC)

That got deleted, and I comment below under "Mean Fitness...". The paragraph doesn't cite any references, so I'm fine with deleting it, but again, this article is about computer science and an optimization technique, not about population biology. A GA can have arbitrarily many sexes; a chromosome can be an SQL database or an OO hierarchy, instead of neucleotides; an individual can be immortal; nerds can date prom queens. It's not biology, it's inspired by biology. Pete St.John 17:21, 15 October 2007 (UTC)

I agree with you that the nature of this article is largely irrelevant to population biology. My only problem with the quote was that it was referring erroneously to traits of natural evolution. If the statement read something like "One significant problem facing the field of GAs is that they maximize individual fitness, when it is generally desired that mean fitness is maximized.", then I would have no problem with it (I'd probably even agree). --SteelSoul 05:00, 16 October 2007 (UTC) —Preceding unsigned comment added by SteelSoul (talk • contribs)

Some problems

This article was a very good overview of GAs. But I thought this sentence was a bit misleading:

The pool is sorted, with those having better fitness (representing better solutions to the problem) ranked at the top.

Although fitness numbers serve to rank individuals, actually sorting them is a very expensive process, rarely done in practice. Roulette and tournament selection were both designed to choose higher ranked individuals stochastically, without the need for explicit sorting. --Bytefield 15:06, 23 July 2005 (UTC)

I'm sorry, but you're completely wrong: while the Roulette and the Tournament indeed are faster than a complete sort, the sort of even thousands of individuals takes microseconds these days, i.e. is completely irrelevant w.r.t. the computation time for the genetic operators and the evaluation function. The real reason for the Roulette or the Tournament is not their speed, but the fact that they do not "sort" the individuals completely: they leave a small chance for underperforming individuals to end up near the top. This is crucial in avoiding premature convergence, as it gives a chance to reproduce even to individuals that seem no good. Why? Well because we do not know whether they are really "no good" (remember, these are combinatorial problems!). -- EF (efalkena@ulb.ac.be)

Yup, plus you don't have to actually sort them. You can select the top K elements in linear time, e.g. [1] In practice, using this deterministic selection process leads to faster convergence which might be good or bad depending on the problem. -- Radu —Preceding unsigned comment added by Raduberinde (talk • contribs) 19:24, 17 July 2008 (UTC)

GAs can rapidly locate good solutions, even for difficult search spaces.

This observation is overly optimistic and general. It should be more carefully worded or more specific. It would be interesting to give an example of a "difficult search space" where GA performs well and alternative methods fail.

Agree completely: GAs are by no means guaranteed to "rapidly locate" anything, even in simple search spaces.
The point is, a GA's encoding and operators must be adapted to the search space (i.e. the cost function) it works with to function properly. It must exploit any inherent structure the cost function has. Failing that, the GA will fail lamentably - but succeding that, it will probably beat any other method on that function. --EF (efalkena@ulb.ac.be)

I think this sentence should be reconsidered:

Unless the fitness function is handled properly, GAs may have a tendency to converge towards local optima rather than the global optimum of the problem.

I wouldn't say the problem of premature convergence is really due to improper 'handling' of the fitness function (what does 'handled properly' mean anyway?). Perhaps someone with more expertise in this subject could elaborate further, in my experience its the genetic operations that can 'handle' this problem and ensure diverse search across the search space.

To go further, one could mention that premature convergence to a local minima is less an issue in GA than it is in other approaches (gradient search, simulating annealing, etc). -- RK

I think I've corrected the first and third problems. I agree with the second one mentioned, but didn't know what to replace it with. — Asbestos | Talk (RFC) 17:12, 23 January 2006 (UTC)

Related techniques

The section on Tabu states that it "moves to the solution with the lowest fitness of those generated". This really sounds wrong and I think it should be "removes the solution with the lowest fitness". I am not an expert but "The enhanced evolutionary tabu search and its application to the quadratic assignment problem" by John F. McLoughlin, III and Walter Cedeño seems to imply this, as well as "The niching method for obtaining global optima and local optima in multimodal functions" by Masako Himeno and Ryutaro Himeno. The both state that the lowest fitness solution is replaced by a new one.

NOTE: I see that User:Diroth changed "lowest fitness" to "lowest energy" here. This sounds more logical. --Jdz 01:31, 25 April 2007 (UTC)

Another problem is that this text is repeated verbatim across several articles, including Ant colony optimization. I think there needs to be a single "List of localized optimization techniques" referenced from GA, AOC, etc. --Jdz 22:14, 21 February 2006 (UTC)

History of genetic algorithms

The history given here credits Holland for introducing genetic algorithms. This is a common misunderstanding. Holland certainly widely publicized these methods among computer scientists, and he did introduce his Schema Theorem. His influence was great. But genetic simulations of evolution had been going on since Barricelli's work of 1954 and Alex Fraser's work of 1957. There were others in that period too: see David Fogel's collection of reprinted papers: 'Evolutionary Computation: The Fossil Record".

In fact by the early 1960's there were many papers in the population genetics literature using simulated evolution, methods that cannot be distinguished from genetic algorithms (they have genotypes, fitnesses, mutation, and recombination). There were even two whole books on how to do genetic algorithms published well before Holland's in 1975:

Fraser, A. S. and D. Burnell. 1970. Computer Models in Genetics. McGraw-Hill, New York.
Crosby, J. 1973. Computer Simulation in Genetics. Wiley, New York.

If Holland invented the method, how do we account for its appearance in those books, and the many papers? Even if one restricts the definition of GA to attempts to solve some design problem outside of biology, Nils Aall Barricelli's work did that too.

I would appreciate any reactions. Perhaps I am missing some restriction of the definition of GA that would rule out Fraser's and Barricelli's work (and all the others), and I want to make sure this is acceptable before I modify the Wikipedia entry. -- Joe Felsenstein (whose 1974 paper on evolution of recombination used simulated evolution, and that too is before Holland)

Go ahead, your contribution is welcome. But do read WP:NOR first—if a Wikipedia entry is based on a "common misunderstanding", that is OK. Wikipedia is a tertiary source, and it aims only to collect and reproduce viewpoints, not correct them. You would need another source than you own expertise for the purposes of verfiability. (See also WP:V). And why not get an account, that makes it easier to talk to you. Best, Arbor 10:46, 25 April 2006 (UTC) Later addition: I think Fogel's work certainly qualifies as the type of source I was talking about. Carry on. Arbor 10:47, 25 April 2006 (UTC)

It is widely cited in academic publications that Holland takes the credit for introducing GA. I guess no matter it is misunderstanding or not, wikipedia should not dispute with academic publications on this topic. Cannot be bothered to attach any reference as in almost every single GA academic publication - books or journal articles, Holland is specifically mentioned for his introduction of GA. - Yin

Holland is no doubt the person most responsible for publicizing GAs. He did not invent simulation of evolution: Barricelli, Fraser and others did. As I wrote, there were two books already in print before Holland's book. Just because so many academic publications get it wrong is no reason for Wikipedia to get it wrong too. Felsenst 14:44, 19 June 2007 (UTC)

The latest editing of this section says that "artificial evolution became a widely recognized optimization method as a result of the work of Ingo Rechenberg in the 1960s and early 1970s - his group was able to solve complex engineering problems through evolution strategies (1971 PhD thesis and resulting 1973 book)." This credits Rechenberg not only with pioneering work on artificial evolution, but specifically says that the wide recognition of the methods is due to his work. What is the evidence that subsequent work cited or was aware of Rechenberg's work? Holland's work was widely cited -- was Rechenberg's? Felsenst 14:50, 19 June 2007 (UTC)

My theory about this dispute is that what first was modelling of biological processes by computer (done by biologists) gradually became using the biological allegory to develop new optimization processes. A computer simulating evolution can be doing about the same thing but with two different goals: to study actual biology, or to omptimize an engineering problem. One is part of the science of biology, the other is part of computer science (applied mathematics, artificial intelligence, control engineering..). Holland was not the first to model biological processes, but he was very early in using similar techniques as part of computer science itself, instead of as a laboratory tool in biology. I don't blame the biologists who pioneeered computer simulations for wanting a share of the credit for GA, but Holland deserves alot of credit for the development of the CS discipline, especially considering how little RAM those guys had in the 60's :-) Pete St.John 16:58, 22 August 2007 (UTC)

I actually agree with this -- there ought to be some way to write the history section to say that biologists such as Barricelli and Fraser pioneered the use the computers to simulate evolution, in the process figuring out how to carry out operations representing mutation, recombination, genetic drift, and natural and artificial selection. Then others, with Holland playing a large role as early and highly visible, used essentially similar methods to solve optimization problems (I think David Fogel's book gives a couple of examples of other early people doing optimization, but Holland's book seems to have caused the Big Bang in using GA for optimization). My insertions of the biologists into the History section was a reaction to the widespread misunderstanding that gives all credit for developing GAs to Holland, as if he were the first person to figure out how to do selection and recombination in a computer. Once I even had a biologist who had been reading about GA come to me and say he had this wonderful idea -- why didn't we use them to model evolution? I had to gently tell him he was 30 years late. BTW I do suspect that Rechenberg is not the person whose work caused artificial evolution to be "widely recognized". Felsenst 05:55, 29 August 2007 (UTC)

I think there is big misunderstanding on why Holland is getting credit. The work he done is formalizing first schema theorem, thus he open the big door for use of GA/EA since one could mathematically show that solutions will converge to "some kind of optima", or at least that fitness of children will increase, until some point, with some probability p. In my opinion this is very important step, thus he deserves big credit for it. Who has first with this idea is not of big importance since they theory had no prove or scientific bases, just hunch that since we have involved one could use this for solving problems. Never they shouldn't by forgotten. GA Fantastic (talk) 08:51, 7 December 2007 (UTC)

"The history given here credits Holland for introducing genetic algorithms. This is a common misunderstanding. Holland certainly widely publicized these methods among computer scientists, and he did introduce his Schema Theorem. His influence was great. But genetic simulations of evolution had been going on since Barricelli's work of 1954 and Alex Fraser's work of 1957. There were others in that period too: see David Fogel's collection of reprinted papers: 'Evolutionary Computation: The Fossil Record"."

bla bla bla

Holland is the grandaddy of GA. You mention books written in the 60's what about david e goldbergs 1953 - ga's in search, optimization and machine learning in which he gives credit to his supervisor john holland. If you have not read that book you should not be commenting on this article. Yin should put his/her google skills in to use somewhere else. —Preceding unsigned comment added by 77.99.241.77 (talk) 14:42, 3 February 2008 (UTC)

I am not Yin but it was I who wrote that quote. Certainly if David Goldberg had written a book in 1953 discussing genetic algorithms, and if this book was a result of a Ph.D. thesis with John Holland, it would mean that the material I added to the page about Barricelli and Fraser would be wrong in crediting them with being pioneers. My (not Yin's) lack of Google skills would be exposed and I should remove the material I added, apologize, and retire from commenting on the GA page. However ... Goldberg's book seems to have been published in 1989. His Ph.D. thesis was actually entitled "Computer-aided gas pipeline operation using genetic algorithms and rule learning." His Ph.D. was granted in 1983, not 1953. Had he been working on it for 30 years? Maybe it was his second Ph.D. with Holland, the earlier having been taken 30 years before? Let's see some evidence. I do agree that Holland was the father of GA, not in having been first but in making these techniques widely known. But he was not first to introduce these methods. Felsenst (talk) 16:01, 3 February 2008 (UTC)

his book was edited in 1989 it was first published in 1953, it's sat infront of me do you want a picture?

Absolutely. Using my (limited) Google skills I find that in 1953 John Holland was 24 years old, with his first publication on GAs not to come until 1962, and David Goldberg would not receive his B.S.E. undergraduate degree until 1975. Our University library's listing do not show him as having a 1953 book, but they do show him as being born in 1953. Most U.S. universities did not have a computer in 1953. Are you sure you're not mixing up his birth date with the date of publication? You can find my email address on my web page which is linked to at my Wikipedia author page. Felsenst (talk) 20:19, 3 February 2008 (UTC)

Yeah citing "Goldberg 53" would be ludicrous. However, the distinction I fail to make convincingly (but I'll keep trying) is between using the biological allegory as an optimization technique vs using a computer to model a biological process. Yes biologists did the later, earlier. That's not the point. Holland was the first (known to me) to do the former. Specifics to the contrary are welcome, if I've missed something. Pete St.John (talk) 21:49, 5 February 2008 (UTC)

I can see the point, and that would argue for greatly reducing the size of the mention of geneticists' work introducing genetic simulations before Holland. However Holland was not the first to use genetic algorithms to solve nonbiological optimization problems. In fact, Barricelli's initial paper had little organisms solving a nonbiological problem. There were others soon after who did this as well -- Fogel in his book "Evolutionary computation: the Fossil Record" reprints these papers (I need to locate my copy). The two efforts are somewhat entangled with each other. Holland certainly should get credit for being the person who publicized the use of GAs for optimization (in the current page Rechenberg is given credit for this), He also introduced the Schema Theorem (although it is often misunderstood as proving that fitness will increase, which it doesn't). Felsenst (talk) 16:16, 9 February 2008 (UTC)

"somewhat entangled" I can buy :-) I just recently stumbled on a lab in Spain where biologists are applying biology to computer science, very interesting. As long as we seek to clarify the entanglement, instead of turf-fighting over precedence, I'm happy for inclusion of (entangled, relevant) genetics material. I'll be interested in the citation to the Fogel and Barricelli's stuff, mabye there's something online somewhere? Thanks, Pete St.John (talk) 20:18, 11 February 2008 (UTC)

David Fogel's 1998 book "Evolutionary Computation: The Fossil Record" is a wonderful collection of many of the earliest papers using genetic algorithms. It includes a number of people who used them very early to solve nonbiological optimization problems. My copy is misplaced but there is available a Table of Contents of the book, which should enable you to track down most of the papers. It is here as Excerpt 3. I can also be contacted by email (my address is on my web site, linked to on my user page). Felsenst (talk) 18:23, 12 February 2008 (UTC)

I'd prefer that people out to establish use of genetic algorithms as an optimization technique prior to Holland, buy expensive books and research their case :-) One challenge of historical analysis is that old books aren't digital and references often aren't digital. I'm not on a campus, so much as I'd be interested by the Barricelli paper (for example) I won't likely dig it up myself. Pete St.John (talk) 18:53, 12 February 2008 (UTC)

That said, the dual appointment you have is very cool; plainly you are in a great position to address the entanglement. But you have graduate students! Put them to work. I believe that even math students should be able to write. Granted biology grad students aren't quite that calibre ;-) Pete St.John (talk) 18:57, 12 February 2008 (UTC)

Observation removed from the main page

GAs cannot effectively solve problems in which there is no way to judge the fitness of an answer other than right/wrong, as there is no way to converge on the solution. These problems are often described as having "needle in a haystack" fitness landscapes.

"A needle in a haystack" problem which is referred here is hard to optimize by any method. Binary fitness values are not enough for a problem to be "a needle in a haystack", in addition it should have only one maximum (a needle). It is easy to think of many cases with binary fitness values that GA can optimize well, simplest ones are line patterns on a 2d greed. Anyways, the observation is not correct as stated.

You're right that the NIAH problems usually refer to those with only one solution, but I think the binary fitness part of the problem is significant. Or, rather, the fitness doesn't need to be binary, but there must be a sharp distinction between correct and incorrect, as this is what makes the optimization difficult (as the GA has no slope which it can climb). Anyway, the point of the statement is to say that "problems with binary fitnesses are difficult," not that "NIAH problems are defined as those problems with binary fitness." I think that this is pretty much agreed upon for basic GAs, as GAs need slopes to climb. What exactly do you mean by the "line patterns on a 2d greed" that are easy to optimize? — Asbestos | Talk (RFC) 17:59, 17 July 2006 (UTC)

I've replaced the sentence, but removed the irrelevent needle-in-a-haystack statement. I've added that a random search can typically find such solutions just as fast as a GA in these situations, as a GA is essentially acting as a large random parallel search algorithm. — Asbestos | Talk (RFC) 13:30, 19 July 2006 (UTC)

Local search ?!

Specifically it falls into the category of local search techniques and is therefore generally an incomplete search.

Um, sure, GAs don't always find the global optimum, but what method does? I can't think much more global method than GA which uses the whole search space as opposed to true local methods which search only the neighborhoods of the given solution candidate. Consider a hill climber local algorithm for example, it halts when it reaches a local optimum. GAs may stall in such position, but in theory they should eventually escape from it.--JyriL ^talk 06:38, 2 August 2006 (UTC)

A method that always finds the global optimum is one that searches every possibility, which a GA doesn't do. I assume that this is the distinction implied by the difference between a global search and a local search. — Asbestos | Talk (RFC) 14:52, 3 August 2006 (UTC)

No, that's the different between complete(?) and incomplete search methods (or do you mean the difference between deterministic and stocastic methods?). Most global optimization methods, including GAs, belong to the latter as the former is usually impossible to implement.--JyriL ^talk 16:37, 3 August 2006 (UTC)

Sorry — I wasn't clear. I wasn't saying that a GA was a local searching algorithm, rather that this was probably what the original author of the sentence had in mind (plus I was answering the "what method does?" rhetorical question). I agree that a GA is not a local searching algorithm. For the record, complete searches are not impossible to implement, in fact they are many times easier to program than a GA (any problem that is solvable by a GA can be solved by a brute-force search). The problem is that they take too long compared to any stochastic method. Anyway: yes, the offending sentence is incorrect, and should go (though the part about it being incomplete is correct). — Asbestos | Talk (RFC) 18:25, 3 August 2006 (UTC)

OK, I was afraid I've made a grand misunderstanding on the issue. ;) "Implement" was wrong word, I meant exactly what you said.--JyriL ^talk 18:30, 3 August 2006 (UTC)

Can someone write the article in simple words please.

How do you mean that please, be more specific ? Tulkolahten 12:08, 16 November 2006 (UTC)

References

The references section has gotten quite long. Many of the references describe applications of GAs or boutique genetic algorithms and aren't cited in the body of the article. I suggest removing the following ones.

Fentress, Sam W (2005), Exaptation as a means of evolving complex solutions, MA Thesis, University of Edinburgh. (pdf)
Harvey, Inman (1992), Species Adaptation Genetic Algorithms: A basis for a continuing SAGA, in 'Toward a Practice of Autonomous Systems: Proceedings of the First European Conference on Artificial Life', F.J. Varela and P. Bourgine (eds.), MIT Press/Bradford Books, Cambridge, MA, pp. 346-354.
Hicks, C., (2006), A Genetic Algorithm tool for optimising cellular or functional layout in the capital goods industry., International Journal of Production Economics, 104(2), 598-614.
Kjellström, G. Network Optimization by Random Variation of component values. Ericsson Technics, vol. 25, no. 3, pp. 133-151, 1969.
Kjellström, G. Optimization of electrical Networks with respect to Tolerance Costs. Ericsson Technics, no. 3, pp. 157-175, 1970.
Kjellström, G. & Taxén, L. Stochastic Optimization in System Design. IEEE Trans. on Circ. and Syst., vol. CAS-28, no. 7, July 1981.
Kjellström, G. On the Efficiency of Gaussian Adaptation. Journal of Optimization Theory and Applications, vol. 71, no. 3, Dec. 1991.
Kjellström, G. & Taxén, L. Gaussian Adaptation, an evolution-based efficient global optimizer; Computational and Applied Mathematics, In, C. Brezinski & U. Kulish (Editors), Elsevier Science Publishers B. V., pp 267-276, 1992.
Kjellström, G. Evolution as a statistical optimization algorithm. Evolutionary Theory 11:105-117 (January, 1996).
Kjellström, G. The evolution in the brain. Applied Mathematics and Computation, 98(2-3):293-300, February, 1999.
Pongcharoen, P., Hicks, C., Braiden, P.M. and Stewardson, D., (2002), Determining Optimum Genetic Algorithm Parameters for Scheduling the Manufacture and Assembly of Complex Products. International Journal of Production Economics, 78(3), 311-322.

Jasper53 07:58, 30 January 2007 (UTC)

My intension is to supply an article about Gaussian adaptation that was used already in 1969 for the maximization of manufacturing yield of signal processing systems. Later it turned out that yield is not very far from the mean fitness of populations of living organisms.

Kjells 16:07, 30 January 2007 (UTC)

I do not feel that the Kjellstrom references will be helpful to someone who is seeking to understand what GAs are and how they work. The references might be more appropriate in an article about the history of evolutionary algorithms.

Text about Gaussian adaptation as a genetic algorithm has been added among GA-Variants, and two referneces to Kjellström has been introduced again. But, of cource, the same references may be found in the article about Gaussian adaptation.

Kjells 20:09, 21 March 2007 (UTC)

I've incorporated some of the references as in-line citations. If other people can help to incorporate the remaining references into the text, then the ones that remain can be considered for removal. -- Bovineone 05:41, 24 June 2007 (UTC)

I apologize for not having included a reference to Rechenberg in the paper about Gaussian adaptation. A reference is available in an earlier paper: Kjellström, G. Evolution as a statistical optimization algorithm. But I am not sure that Rechenberg used Gaussian adaptation, because it relies on a certain theorem that was not discovered by Rechenberg, I suppose. In addition, focus is on maximization of mean fitness and average information, not solely the fitness of the individual.--Kjells 06:37, 25 June 2007 (UTC)

If I remember rightly, the probability of success in Rechenbergs algorithm was equal to 0.18 instead of 0.37 as it should be in Gaussian adaptation. The reason is that Rechenberg maximizes the fitness of the individual instead of mean fitness and average information. This follows from the theorem of efficiency as published in Kjellström, G. On the efficiency of Gaussian adaptation. So, I can't see that he used Gaussian adaptation at that time.--Kjells 13:09, 25 June 2007 (UTC)

Some lines about Rechenberg have now been included in the article on Gaussian adaptation. And I don't think that Rechenberg was prior to Kjellström with Gaussian adaptation, and I think that the corresponding line should be removed--Kjells 06:19, 26 June 2007 (UTC)

GA's now important in statistical experimental design

Dfarrar 22:15, 21 March 2007 (UTC)

Dfarrar, you created a section title but no section? Pete St.John 15:24, 22 March 2007 (UTC)

Rechenberg and Gaussian adaptation

Gaussian adaptation is about maximization of manufacturing yield and tolerancing. Later this has been referred to as centering and tolerancing. As a model of evolution Gaussian adaptation is said to carry out a simultaneous maximization of mean fitness and average information by the aid of a certain theorem of Gaussian adaptation, not discovered by Rechenberg.

After having searched the litterature in these contexts I can't see that Rechenberg has anything to do with Gaussian adaptation. He has not been referenced in papers about centering and tolerancing. I have therefore removed the lines in the article about Rechenbeg and Gaussian adaptation.

Evolution strategy is included as a variant of a genetic algorithm discovered by Rechenberg.--Kjells 06:32, 27 June 2007 (UTC)

After the latest changes the reference to the first usage of Gaussian adaptation has disappeared. It has now been included again.--Kjells 18:31, 9 July 2007 (UTC)

Inversion operator

There has been a discussion on the inversion operator on my talk page. I think it would be better to describe this as a rule of genetic reordering variation among other mutations rather than by a single word in the text.--Kjells 09:45, 24 July 2007 (UTC)

Examples

I think it could be useful to add a link or two to some genetic-algorithm applications or evolution movie (for example: a walker). In this way it should be more clear for users how ga work...

I have added a link (in the section external links) to a tutorial is presented step by step a genetic algorithm. I think it is usefull for people that after reading the theorie would like to programm one. It could be seen as an example done by the reader itself. The link is http://fog.neopages.org/helloworldgeneticalgorithms.php. Swarmcode 12:47, 17 August 2007 (UTC)

yeah and it got removed (once anyway). It's really not so bad, but the English can be improved. Keep working on it, be persistent and be patient :-) My $0.002 worth of advice. Pete St.John 15:59, 17 August 2007 (UTC)

Hey, I agree this site is pretty good, and useful if the English was a little better, but it doesn't meet the criteria of the wikipedia project. Ultimately that is what is important. I would love to link to my homepage and dissertation on GA's, but I'm not about to! That is what a userpage is for, which I'm pleased to see Swarmcode has now got, :D welcome! MattOates (Ulti) 19:23, 17 August 2007 (UTC)

Following the link Swarmcode just gave above, I realize that you have to click on the DNA looking “HelloWorld” image to get to that article. I didn't even figure that out from the original link! My question is: would anyone else notice? Would they see this as a well-referenced & authoritative source, worth being directly linked from the wikipedia article compared to any other website on the net? If the answer to these is honestly no, should it be linked? Sorry if I sound like an ass (I know that I do :[ ) I'm just tired of following external links to find they aren't as informative as I would hope. Quick P.S. Swarmcode, you might like to contribute to Wikiversity with that material, since it is way more in keeping with the content they are responsible for. MattOates (Ulti) 19:44, 17 August 2007 (UTC)

Well, my site is not a well-referenced because that is plenty of other sites out there that are well referenced. It is about practical things... how to implement and although I have looked for some practical tutorials about GA I have not found out there and in this way my site is unique (the user has a code in the end to play with). Check out Nehe and you will see why their tutorials are so easy to understand and learn. They did not have reference, but did he need? If somebody want some reference they could easily find it in a book or papers, or even in the 2 cites cited in the external links of wikipedia. My tutorial and answer here goes to where could someone learn something more practical (in wikipedia for example).. something to run and compile right after the tutorial not only some letters. If you dont wanna link my tutorial that is ok ... but as a user of wikipedia I am very sad about not linking with sites that give Practical things. Not that theorie is not good, but they are not enough.... I think the first guy who posted the question had the same sensation as myself.Swarmcode 00:16, 22 August 2007 (UTC)

What about Genetic Arm? It should be a good example... —Preceding unsigned comment added by 146.133.1.43 (talk) 14:16, 28 December 2007 (UTC)

Mean fitness, and biology vs computer science

Steelsoul reverted the paragraph "...A problem that seems to be overlooked by GA-algorithms thus far is that the natural evolution maximizes mean fitness rather than the fitness of the individual (the criterion function used in most applications)...", reason given that (paraphrased) "consensus in population biology contradicts". I let the revert stand because the item did not cite any references. However, in general, I hope we remember that this is about computer science, not population biology. We should not draw inferences from one field about the other, without references. This has been a source of confusion in the article, particularly the historical material, where early models of population biology have been confused with early biology-inspired optimization techniques, in arguing over precedence. Pete St.John 17:14, 15 October 2007 (UTC)

Proportion of individuals is/are

I reverted this sentence: "Most functions are stochastic and designed so that a small proportion of less fit solutions is selected..." to "...are..." (italics mine). First, if we swap back and forth between Americanisms ("are") and Britishisms ("is") then we will wear ourselves out. I advocate tolerating each other's regionalisms, although it will be a good thing when we all converge. Second, however, in this case the Americanism is a bit more precise, because in the algorithm the proportion is a parameter and the individuals (plural) are individually selected. "Selection" refers to the individuals, not the proportion; but the proportion of such individuals (of less fitness) is small. So IMO the original statement is slightly better. Pete St.John 16:13, 22 October 2007 (UTC)

Applications

I notice that there is a rather long list of applications... I think we should be sure that those are really applications for which genetic algorithms have been researched, and possibly even have shown some degree of success.

For instance, I am a bit surprised at the mention of code-breaking, as it is a typical case where 1) there is no possibility to ascertain the quality of a solution, 2) A "close to optimal" solution is worthless.

The list is slightly too long already, and if we don't shorten it and find references, we might just add to it "finding the answer to life, the universe and everything" Ratfox 08:35, 13 November 2007 (UTC)

I'm gonna revert "hardware bug finding during validation" and post to the contributors talk page asking for a citation. Pete St.John (talk) 17:49, 26 November 2007 (UTC)

Belongingness to Molecular and Cellular Biology WikiProject.

Hi! I was near chock experience then I saw that this page was added to Molecular and Cellular Biology WikiProject. I don't agree and don't understand why would someone do this? This is the topic, with no doubt, belonging to Computer Science. so it should by removed from Biology Project.

Regards GA Fantastic (talk) 08:56, 7 December 2007 (UTC)

I agree, it shocked me too for the same reason. Part of the explanation may be the claims made by biologists that early computer models of genetic processes were precursors to genetic algorithms; it's a credit dispute (see above). As computer science, I think we pretty much agree that Holland was very early and deserves alot of credit (to me, it's amazing to try and do GA with such tiny amounts of memory) but we can also agree that as a model of a biological process, GA built on work done by biologists (including computer simulations). I don't so much mind including this in a bio project, but we should be in AI and CIS projects also. Pete St.John (talk) 19:49, 7 December 2007 (UTC)

That said, I see the list of categories on the article page is very reasonable. One might add "AI" but "search algorithms" and "evolutionary algorithms" are already subcategories of AI. I'll go drop a note on the cellular bio project page and see if they still care. Pete St.John (talk) 19:57, 7 December 2007 (UTC)

so, I just dropped a (lengthy) note about it at Wikipedia_talk:WikiProject_Molecular_and_Cellular_Biology#Miscategorization_of_Genetic_Algorithms Pete St.John (talk) 20:24, 7 December 2007 (UTC)

I added the Wiki project Computer Science template, so even if people are confused, they will know they are confused :-)

Well, I'm the one who put the MCB template on here way back when I was tagging hundreds of articles. The reason I tagged it was because GAs have their application in various aspects of computational molecular biology. I found other bioinformatics algorithms, e.g. BLAST, were tagged with the MCB template as well. In computational molecular biology, GAs are used for multiple sequence alignments and phylogenetic tree building, for example. Therefore, I felt this article was within the scope of bioinformatics/computational molecular biology. - tameeria (talk) 22:59, 7 December 2007 (UTC)

I like to say that GA is never the best solution, but always an adequate one, so it's good for problems that lack well-developed theoretical frameworks. However that may be, GA has lots of applications. I have used it for Extremal Graph Theory-- and in fact, an algorithm is a tree, which is a kind of graph, so really GA should be in the Graph Theory project :-) The interconnections are diverse, varied, and numerous. If you take a gander at the list of applications you will see many fields; it would not be helpful to insert GA into every such category. I hate to be a "Streamliner" but I vote to drop the MCB category. For example, you can find GA in the course catalogs at CS departments, but not Bio, correct? We all use computer techniques, from Word Processing to Nonlinear Optimization, but those things aren't necessarily parts of our fields of appplication. Now Genetic Computing, that would be tuff to categorize. That, I think, would be both CIS and MCB (well, not cellular but molecular). And using a Genetic Computer to run a Genetic Algorithm to optimize the fit of a Model of Genetics would be...confusing :-) Also I think I should add that GA, which we mostly think of as Nonlinear Optimization (I think of it as AI), predates Bioinformatics, and I don't think particularly pertains to it. I'm sure they use lots of algorithms in Bioinformatics. Pete St.John (talk) 23:30, 7 December 2007 (UTC)

GA is at least mentioned if not covered in some detail in bioinformatics classes taught to biology students. These students may then come here to learn more about e.g. how it works, what it's used for, and how it differs from other search algorithms. Biologists with an interest in sequence alignments, phylogenetics, or structure and pattern analysis may come here. And yes, sure, they use lots of different algorithms and most of them will be on the more or less clueless "user" rather than "programmer" side. The question that might bring a biologist to this page might be whether GA would be an appropriate choice for their search problem and what advantages/disadvantages it might have over the other options. I'm not sure that's clear from this page at all.

Also, Wikiprojects are not exclusive but there can be a whole list of them collaborating for any page so I'm not sure "belongingness" would be an issue. Many articles have a whole list of Wikiprojects collaborating on them. There is no doubt that this subject is part of computer science, but the term "genetic algorithm" also brings up 1372 publications in PubMed indicating its importance for the life sciences as well. - tameeria (talk) 02:47, 8 December 2007 (UTC)

Your are making a god point here, but I would like to propose more correct model. I think it would by handy if pages could by tagged as "is of importance for" or "is of interest for". This would clear all confusions. I think articles should have clear and straight belongingness. Then if they are of interest for particular groups they could by tagged as so. Use of GA can by found in physics, expersystems, physical chemistry, crystallography, math and so on. Also a lot of computer science is of big interest for many different science branches. Adding those to different wiki project wouldn't be helpful, only confusing. Thank you.

GA Fantastic (talk) 07:33, 8 December 2007 (UTC)

There are many interdisciplinary topics on Wikipedia and it's often impossible to sort them into a single category. For example, look at oxygen - that article has tags for three Wikiprojects: Elements, MCB, and Medicine. The tags indicate which Wikiprojects consider oxygen as being of importance to them. I thought the Wikiproject templates were just that: indications of what is of interest for a particular project. Regarding importance, there's actually a rating feature built in. So I've rated this one low importance for MCB because quite frankly I think there will only be a small subgroup of molecular biologists really interested in this. If you want to rate the importance for the Computer Science project as high, I think that would help solve the confusion about for which project the topic is more important. I've rated it's class as B for the MCB project. It's a good article, but it is obvious is was written neither by nor for biologists, so it could be improved with some information putting it into the context of bioinformatics maybe. Again, your rating may vary if you think the article is a complete coverage of the topic from the computer science side. - tameeria (talk) 15:26, 8 December 2007 (UTC)

I'm thinking it's mainly the difference between a method of study and an object of study. We all can use GA as a tool in our study, but GA is an object of study in CS. Biology papers would be about using GA to learn something about Cells, while CS papers would be about using Mutation Rate for better GA. Ironically, my own interest is using GA to evolve better GA, so (in this activity) it's both the tool and the object. Pete St.John (talk) 21:05, 11 December 2007 (UTC)

In the meantime the article had been moved from WP:WPMCB to Wikipedia:WikiProject Mathematical and Computational Biology. I don't agree with this for the same reasons as for MCB. While computational biology uses computational methods to solve biological problems, while GA is a biologically inspired methodology in computer science. Thus, I've removed the banner. Hendrik Fuß (talk) 13:45, 17 January 2008 (UTC)

I would myself just have the one category, but the WPMCB folks seem to like it, I think partly on historical grounds and partly from ongoing crossdevelopment from using GA to model some genetics processes. So I was content to have both categories CS and Computational Bio (which was a better fit than MCB); at worst, it would seem to do no harm. The MCB cat I think could confuse people unfamiliar with the subject. Pete St.John (talk) 19:17, 17 January 2008 (UTC)

"Belongingness" revisited

I just wanted to point out an interesting reference: Natural Computing Group (at Politecnica de Madrid, but the site is in English) is a group of biologists (sorta) working on applying biology (in more than one way) to Computer Science (as opposed to using computers to model biology; everyone uses computers to model what they do). So for example:

More recently, two clearly biologically-inspired paradigms are being applied very effectively to solve applied problems in business and industry: artificial neural networks and genetic algorithms. However, there is still room for advancement. Instead of developing computing systems inspired by biological processes, biological substrates and biological processes can be used directly to codify, store and handle information, as Leonard Adelman demonstrated with his pioneering experiment on DNA computing.

So I think I was too dismissive of the Molecular-Cellular Bio categorization, but of course the Mathematical-Compuational Bio cat is a better fit. (Funny they both have the acronym MCB.) Pete St.John (talk) 15:46, 31 January 2008 (UTC)

Exact vs approximate solutions

In a discrete system (such as optimizing graphs with large numbers of nodes) exact solutions can be found with GA. So I'll considering reverting that change just now. Pete St.John (talk) 20:24, 7 December 2007 (UTC)

Genetic memory deletion

The article genetic memory is currently up for deletion. I've done some literature digging and found that the term is also used in computer science publications in connection with genetic algorithms. So just a heads-up in case anyone wants to comment on the delete discussion. - tameeria (talk) 15:47, 12 December 2007 (UTC)

Addendum: The discussion has been closed and the result was to keep the article with the suggestion to do a major rewrite, so in case anyone from the computer side wants to contribute by adding/rewriting please feel free to do so. - tameeria (talk) 19:06, 12 December 2007 (UTC)

Pseudo-code algorithm

is this correct? The link below shows a more accurate pseudo code in my opinion.

http://www.obitko.com/tutorials/genetic-algorithms/ga-basic-description.php —Preceding unsigned comment added by 77.99.241.77 (talk) 14:49, 3 February 2008 (UTC)

application proliferation

A challenge we have with this article is the very large number of applications and dissertations which can be linked here (generally by their authors) but which can collectively bloat the article. In the recent case, I reverted such an addition just because it introduced a typographical error. I propose we insist that if someone wants to link his thesis, he at least do a good job of it :-) But seriously, theses that help us understand the general subject are welcome. Otherwise, they may be more pertinent in articles about the application rather than the method. Pete St.John (talk) 19:31, 25 February 2008 (UTC)

Cat: algorithms

Andreas deleted the category "algorithms"; I'll go ask why at his talk. I would agree that GA hasn't gotten much attention from complexity theorists (but maybe should). But surely they are algorithms. Pete St.John (talk) 22:01, 25 February 2008 (UTC)

This article is already listed in categories "Evolutionary algorithms", "Genetic algorithms", "Optimization algorithms" and "Search algorithms". All of these 4 categories are subcategories of "Algortihms" category. The category "Algortihms" should list very few article pages directly and should mainly contain subcategories. This is why I thought that this article don't need to have "Algortihms" category. Andreas Kaufmann (talk) 22:33, 25 February 2008 (UTC)

That makes sense. If in general Algorithms categories are getting hierarchically structured, that's great, and I don't want to interfere with the people doing that good work. So please go ahead. Pete St.John (talk) 18:20, 26 February 2008 (UTC)

schemata vs cylinder sets

After a quick skim of the concept of Holland schemata, I get the impression that these are one and the same, identical to what the rest of the world calls cylinder sets. Specifically, for example, for binary strings of length 6 e.g; such as the "schemata" 0**1*1, this would be called an open (cylinder) set on the product topology for 2^6. Is there any way in which a schemata is anything other than a cylinder set in the product topology? linas (talk) 21:42, 5 June 2008 (UTC)

No Mention of the Schema Theorem??

First of all I want to say right up front that this article should not be nominated for anything. It does not clarify this topic at all, but only makes it seems more obscure. I have several things to say here. First of all, we have these various phrases floating around whose meanings should be layed to rest once and for all by wikipedia. Among these are Evolutionary Algorithms, Genetic Algorithms, Genetic Programming, Evolutionary Strategies. We should emphasize that GAs use a phenotype and a genotype in each individual, and that the existence of the genotype is the defining aspect of a GA. The mentioning of a genotype and a phenotype needs to happen right at the top of this article.

Genetic Programming is completely different in that it evolves literally, code strings. There is no aspect of the expresssion of a genotype into a phenotype in Genetic Programming. Therefore it is not the case that Genetic Programming is a "subset" of GAs. Genetic Programming is the most badly-named discipline in computer science!

On the subject of premature convergence of the population into a local maxima : This issue should be isolated and described all by itself in a separate section. As it stands now, this subject is hinted at in various parts of the entire article. The problems of early convergence and exploitation-vs-exploration does not need to be mentioned half-way in a wikipedia article. It can be isolated and talked about in depth. I think this article needs a section titled Exploitation Versus Exploration. And within this section should be all the various descriptions of mutation rates and crossover rates and other related material. It is not a sufficient explanation to say "depending on what the fitness landscape is like blah blah blah". How does suddenly referring to a fitness landscape make this subject more clear to a layman reading the article? Think about it.

Where is the Schema Theorem? I'm sort of shocked that someone could write an entire encyclopedia article on Genetic Algorithms and never even mention the schema theorem. The schema theorem is basically the mathematical justification for genetic algorithms. Someone might come away from this article thinking that Genetic Algorithms are willy-nilly theory. There are two ways to go about this depending on style. One way is to make a section like ---Mathematical Justification--- or perhaps something like ---Schema Theorem--- On a personal note, I am taking offense that this article's claim that mutation is the only thing driving a GA, and that crossover is just a "fancy form of mutation". This is not a debating forum, this is an encyclopedia, therefore we are obligated to write only things we know for sure. paros (talk) 05:01, 6 June 2008 (UTC)

Do you mean Holland's schema theorem? I myself am shocked to observe that the russian work on the renormalization group, which "explains" and gives mathematical credence to this whole subject, is not mentioned. Perhaps its too mathematically abstract, and hard to understand? As to poor article structure, this is typical of many multi-author wikipedia articles. It takes real work to straighten all of it out. linas (talk) 14:42, 6 June 2008 (UTC)[

This article has been significantly improved in the last 9 months. Cheers! Miloserdia (talk) 02:57, 1 April 2009 (UTC)

Merge Quality control and genetic algorithms here

I would like to ptopose to merge the Quality control and genetic algorithms here, because that combined article seems to fail notablity. -- Marcel Douwe Dekker (talk) 15:39, 18 May 2009 (UTC)

Building block hypothesis

This section may need attention. The quote appears to suggest that crossover is used in genetic algorithms to generate variation.

My understanding is that the function of crossover (sexual reproduction) in nature is to cross in sections of genetic code with mutations resulting in advantageous variations and cross out corresponding disadvantageous variations. By recombining two copies of a gene via crossover, the genetic carrier is made more robust to random variation which might result in corrupt code (since both copies would need to be corrupt in the same place to completely corrupt the genetic code). Thus crossover is in fact a way of limiting the effect of variation through mutation and collecting random variations which have increased fitness into a single sequence of genetic code, not a means of generating variation in that code.

I added a link here to Holland's Schema Theorem. I think the author's meaning of this section is to give a hypothesis for why GAs work. The GA uses 'building blocks' or schemas (high fitness but undefined parts of a solution) to 'build' the optimal solution. These schemas then become more frequent in each subsequent generation. The article gives the formula. To make it more complete though, you'd have to dive into literature to find out what is written on the matter, starting with Holland and Goldberg, I guess. 145.53.202.88 11:13, 16 November 2006 (UTC)

It is a bit problematic this wikipedia entry. You are talking about BBs here and thus I suppose the focus should shift from the sGA to PMBGAs or other algorithms that use the BBs. And if you are including PMBGAs in this section, it is important to notice that complex problems such as deceptive problems can be solved. —Preceding unsigned comment added by 201.63.228.171 (talk) 23:09, 23 February 2010 (UTC)

Reversion because of alleged copyright refringement

Dear Oli,

I followed wikipedia's copyright policy, but could not see which part of my revision refringed copyright. It must be very easy for you to spot - i wonder if you'd be kind enough to spare a minute to point out which part to me, please? Appreciated.

The author of the page http://userweb.elec.gla.ac.uk/y/yunli/ga_demo/ has donated its copyright:

"Copyleft: The text of this webpage is available for modification and reuse under the terms of the Creative Commons Attribution-Sharealike 3.0 Unported License and the GNU Free Documentation License." ieee (talk) 20:07, 17 January 2010 (UTC)

Glasgow Genetic Algorithm Demonstrator - Learning wiki

I like wiki and have found it's most useful when I need to pull information fast on something that I don’t know. While I do not claim no coi, the purpose of adding that GA demo link was obvious to me – to help newcomers to EC/EA/GA learn quickly, and many who were introduced to the link did learn fast from it. However, all seven tutorials there cannot do the job as efficiently or effectively as this one. What made you suggest that any of the seven were better than this one?

I'm not claiming I'm an expert, but do have practiced GA for many years. I understand that several universities use it in their teaching, eg:

University of Glasgow of course – where the link is originated: http://userweb.elec.gla.ac.uk/y/yunli/ga_demo/

University of Sussex: http://www.informatics.sussex.ac.uk/courses/nsai/

University of Granada: http://decsai.ugr.es/~ocordon/Docto-CE-RI-09-10.html Demo de Funcionamiento General de Algoritmos Genéticos

Other places, like ai-depot: http://geneticalgorithms.ai-depot.com/Applications.html

Since wiki is contributed and edited by the users, can't we let the users decide which one is better? Is there a way for wiki to monitor the hit rate or something that is more objective?

I believe that we both would like to see wiki shines with its user involvement and with its wider and quicker availability than other encyclopedias. ieee (talk) 22:00, 21 January 2010 (UTC)

Then we should move this discussion to the article talk page so it's easier to get others' involved. --Ronz (talk) 22:11, 21 January 2010 (UTC)

There are seven other links listed under "Tutorials." That's too many. One or two should be enough, and need to meet WP:EL.

I think the amount and detail of information provided in the seven other links are to the one I've removed. --Ronz (talk) 00:55, 23 January 2010 (UTC)

Hmm, this is an electronic encyclopaedia and thus electronic/interactive tutorials should shine. Please do test drive every entry and see which would provide fastest help at this electronic age.

PDFs are similar to a normal book, but even having said this, I do not mind pdf files there because they're more searchable, still better than the index page in a hardcopy encyclopaedia.

ieee (talk) 01:17, 23 January 2010 (UTC)

I've removed it, given the promotional material added about it [2]. Seek outside perspectives through WP:THIRD or WP:ELN. --Ronz (talk) 16:02, 30 January 2010 (UTC)--Ronz (talk) 16:02, 30 January 2010 (UTC)

Bibliography

The bibliography section seems to be a hodge-podge collection of arbitrary sources with no obvious connection to the article text (i.e. no inline citations). Can we improve the situation? If not, I'll delete the bibliography (after which sources should be added back with accompanying inline cites). Oli Filth^{(talk|contribs)} 17:01, 17 June 2010 (UTC)

Agreed - maybe just do the latter and start putting in cn's where necessary? I am a little sick and tired of finding this sort of thing all over wikipedia. It seems to have stemmed from articles written over three years ago that haven't been touched much since. Chaosdruid (talk) 17:14, 24 July 2010 (UTC)

Referencing

Hello,

I popped over here, because I noticed User:93.96.195.99 putting in several citations to Li and friends, which I was investigating to see if there was undue weight. Whilst reading the "cited" section, I note that there is an entire section which is a WP:LIST, which is cited very well, but with none of the citations asserting any kind of fact, which is unusual. The normal method of citation is to state a fact, then cite a source that supports the fact (see WP:CITE). I have edited in the spirit of WP:BOLD (too bold?) and removed the section. User A1 (talk) 20:30, 9 August 2010 (UTC)

Hi

You have deleted all of the applications of GA ? Hmmm...I don't know how others will react but I am fairly amazed that you did not discuss first or consider moving them to a "List of..."

I will do that now as I have checked and seen that the list of applications has been in there for over two years and added to over time and has not been an issue before.

If it is decided to restore then so be it...

Chaosdruid (talk) 21:40, 9 August 2010 (UTC)

Hans Bremermann link

Should the empty Hans Bremermann link not be Hans-Joachim_Bremermann? Or, alternatively, redirected somehow? Robbiemorrison (talk) 06:51, 14 October 2010 (UTC)

Berkely links German-born Hans Joachim Bremermann to GA; see http://berkeley.edu/news/media/releases/96legacy/releases.96/14319.html. Glrx (talk) 16:45, 14 October 2010 (UTC)

Should all reference to the Generative Fixation Hypothesis be removed?

Should the following statement be removed from the end of the "Criticism of the building block hypothesis" section of the article on Genetic Algorithms? 16:24, 24 June 2010 (UTC)

"Recently, a new explanation---the Generative Fixation Hypothesis---has been proposed.^[1]"

A transcript of the dispute can be found on Oli Filth's talk page

To summarise the "remove" argument :

This was added as a COI edit; linking to one's own PhD thesis almost never acceptable!
There is no indication that this theory has had any adoption in the field; it's less than a year old, and has had no citations (so far as Google Scholar will tell us).
Until this theory does become notable, it would be inappropriate to discuss it in the article.

The "keep" argument:

The generative fixation hypothesis (GFH) is one of only two published, peer reviewed, non-folk explanations for the adaptive capacity of simple genetic algorithms (the other being the building block hypothesis).
The generative fixation hypothesis is the only theory to date that claims to account for the adaptive capacity of simple genetic algorithms with uniform crossover (UGAs).
In practice, UGAs frequently outperform genetic algorithms with one point crossover, even though according to the building block hypothesis, they shouldn't. The GFH explains why. In doing so, the latter hypothesis addresses an anomaly that has been outstanding for over twenty years.
Given the above, a domain expert writing from a neutral point of view about genetic algorithms would be remiss if she did not mention at least the existence of the GFH; unless, of course, she finds a fatal flaw in the GFH that was missed during peer review.
By his own admission, Oli Filth knows next to nothing about genetic algorithms. He is, therefore, not in a position to determine if a domain expert writing from a neutral point of view would or would not make the statement in question. Keki Burjorjee (talk) 16:24, 24 June 2010 (UTC)

^ Burjorjee, Keki M. (2009). Generative Fixation: A Unified Explanation for the Adaptive Capacity of Simple Recombinative Genetic Algorithms (PhD thesis). Brandeis University. OCLC 500907878.

Oli Filth's Talk Page

I'm inserting a copy of Oli Filth's talk page. Glrx (talk) 17:42, 28 July 2010 (UTC)

Dear Oli, I see that you've removed my recent edits to the genetic algorithms article on the grounds that it is "self promotion"? I'd like to get a sense for what you mean by this. Is it automatically "self promotion" when an author writes about his published work on Wikipedia? The generative fixation hypothesis is one of just two non-folk-hypotheses about the adaptive capacity of genetic algorithms (the other being the building block hypothesis). There is indeed a large body of "theory" about genetic algorithms; but, this work is best understood as "theoretical approaches to genetic algorithms"; apart from work related to the BBH and the GFH, none of it has, to date, produced a plausible explanation for the adaptive capacity of genetic algorithms. Another question: Eric Haugen writes above that the generative fixation hypothesis is in dispute. What are his sources? Keki Burjorjee (talk) 16:47, 14 June 2010 (UTC)

In general, editors who cite themselves (in particular, in cases of very new publications) are doing it purely as self-promotion. If that was not your intention, then I apologise for jumping to conclusions, but it's a very common pattern here at Wikipedia. If your work is notable (I have no idea; I don't claim to be anything close to a subject expert), then I'd prefer to see it referenced by other editors, because otherwise it raises questions of conflict of interest. I'd also prefer to see evidence of wider notability (i.e. has the notion of "generative fixation hypothesis" been picked up by others in the field? have your papers been cited?, etc.) Oli Filth^{(talk|contribs)} 16:53, 14 June 2010 (UTC)

I understand your reservations. I believe that the generative fixation hypothesis is notable by virtue of being one of only two published, peer-reviewed non-folk explanations for the the adaptive capacity of genetic algorithms. It is also the only published, peer-reviewed explanation for the adaptive capacity of simple genetic algorithms with uniform crossover. —Preceding unsigned comment added by Keki Burjorjee (talk • contribs) 17:21, 14 June 2010 (UTC)

It would be nice to have this conversation somewhere else. Anyway, I disagree that it's notable because there are only 2 theories. It's notable when other reliable sources refer to it, etc. To be clear, I am not weighing in on whether it is notable; I am just discussing the criteria. ErikHaugen (talk) 17:32, 14 June 2010 (UTC)

Erik, that's certainly one criterion. But is it the only one? When evaluating the notability of a new, published, peer-reviewed theory, I believe that some additional points should also be taken into consideration---especially in a proto-scientific field like genetic algorithmics: i) the degree to which the new theory diverges from the prevailing theory, and ii) the potential impact on a field if the new theory turns out to hold water. Scientists sometimes take time to digest fundamentally new ideas. So, a new theory might not be cited immediately after its publication. Does that automatically make the new theory non-notable? The Generative Fixation Hypothesis is the first theory since the 1970s to give the building block hypothesis a run for its money. And, as I've previously mentioned, it is the only plausible explanation that has ever been proposed for the adaptive capacity of genetic algorithms with uniform crossover. I believe that removing all mention of this hypothesis from the genetic algorithm article does curious, seeking readers a disservice. Keki Burjorjee (talk) 04:58, 15 June 2010 (UTC)

Please see WP:NOTABILITY for the Wikipedia policy on notability. We can't just invent new criteria for protoscience. If this theory does have an impact on the field at some point in the future, then it will be cited, and then it will fulfil the criteria for notability. Until then, there's no need to mention it in the article. Doing otherwise really would be doing readers a disservice, because it would give a false impression of the notability of a particularly nascent school of thought. Please see WP:UNDUE. Oli Filth^{(talk|contribs)} 08:17, 15 June 2010 (UTC)

I concede that my most recent edits may (kinda sorta) give the impression that the GFH is "equal" to the BBH in terms of adoption. Giving readers this impression was not my intention. I sought only to inform readers that there is an alternative to the BBH (a theory whose failings are well known), and to give a brief overview of the GFH. Nevertheless, I can see why you might view these changes as "overreaching". Would you be okay with rolling the article back to the point before my most recent edits? Another option is to let my edits stand, and to add a banner to the section on the GFH that cautions readers that the section contains material that may violate WP:NOTABILITY and WP:UNDUE. Keki Burjorjee (talk) 16:38, 15 June 2010 (UTC)

Well, the latter is certainly not an option! And to be honest, I'm not particularly keen on the former idea either. As far as I'm aware, your theories were published a little over 6 months ago, and have not yet been cited or discussed by any independent authors. The fact that you had to add them to Wikipedia yourself is indicative that they have not reached even the lowest rung of notability yet. This is not meant to be a criticism or a putdown; I'm merely pointing out that with the available evidence, there is no compelling reason to mention your theories at Genetic algorithm yet.

I would suggest that if you wish to continue to debate the issue, then bring it up at Talk:Genetic algorithm, so that a wider array of editors may air their opinions. Oli Filth^{(talk|contribs)} 17:32, 15 June 2010 (UTC)

Six months eh? If only scientists were that reliable ;-)

I think you forget that you and Erik aren't the only policemen on this beat. There are others with domain knowledge greater than yours who would have removed, or contested my statements, and the reference to my dissertation if these were indeed unhelpful, or misleading. (My statements are provocative enough, and so is my dissertation) That neither have been removed, or contested for months should tell you something. In any case, I've raised the issue on the Talk:Genetic algorithm page as you suggested. I don't know how many people actually read that page without being prompted. Is there a formal procedure for getting some third party attention? Keki Burjorjee (talk) 06:39, 16 June 2010 (UTC)

Yes, see WP:3. However, it says "If more than two editors are involved, 3O is not appropriate." So I suggest you read through WP:DISPUTE to find a different approach.

By the way, don't be so sure that "no-one removed my edits" == "my edits were fine"! All sorts of inappropriate material (nonsense, vandalism, self-promotion, etc.) sneaks its way into Wikipedia articles all the time, and often sits around for months or years before anyone notices. The people that read the talk page are typically the people that "police" the corresponding article page, so if your guess that perhaps no-one will read that talk page without prompting is correct, that could also explain why some of your edits stuck around for a while. Oli Filth^{(talk|contribs)} 08:48, 16 June 2010 (UTC)

Are you sure about that? I've done some deleting on Genetic algorithm myself (not enough, unfortunately), and can't say I've spent much time on the talk page. Thanks for the links. Keki Burjorjee (talk) 03:49, 17 June 2010 (UTC)

I did some digging. It turns out that your application of WP:Notability is incorrect. The nutshell box at the top of that page says "Notability does not directly affect the content of articles, but only their existence". I am not trying to set up a wikipedia page devoted to the generative fixation hypothesis; so, WP:Notability does not apply. It also turns out that a Ph.D. dissertation available on UMI (mine is) is considered a reliable source because it has been vetted by a committee of scholars. See the subsection on Scholarship in [3]. Up next, Conflict Of Interest. On Talk:Genetic algorithm you write "This was added as a COI edit; linking to one's own PhD thesis almost never acceptable!". You assume, in other words, that writing about one's own work is almost-by-definition a COI edit. The following sentence from Wikipedia:No_original_research#Citing_oneself says otherwise: "If an editor has published the results of his or her research in a reliable publication, the editor may cite that source while writing in the third person and complying with our neutrality policy". As mentioned previously, Ph.D. dissertations available on UMI count as reliable publications, and I do believe that I've abided by Wikipedia's neutrality policy. As mentioned previously, I agree that my most recent edits fall short with respect to WP:UNDUE. I'll figure out how to handle this better. In the meantime, I'd like to roll the section on the building block hypothesis back to its state in the following revision [4]. Any objections? Keki Burjorjee (talk) 07:06, 17 June 2010 (UTC)

In response:

You are correct about WP:N; however the criteria it sets out are, in practice, typically used as a benchmark for article content as well. If we didn't do this, it would be a free-for-all of people adding mentions to all sorts of non-notable research.
I never claimed that it wasn't a reliable source!
There's a difference between citing your own work to back up existing article content, and adding a section to an article about your own work. This is a case of the latter, and that is a conflict-of-interest. With respect, the fact that you are now fighting tooth-and-nail to get this reinstated is a further indication of COI.
As I've already said, unless you get some kind of consensus (basically at the thread you've set up at Talk:Genetic algorithm), there's no justification for adding this material back. If your work is truly worth mentioning at the moment, then let another editor bring it to the article (that's the spirit of WP:COI). Oli Filth^{(talk|contribs)} 08:39, 17 June 2010 (UTC)

In response to your first point: You can't have it both ways Oli. Either we are both bound by Wikipedia's written policy, or neither one of us is. Whether or not there would indeed be a free-for-all is beyond the scope of the present discussion. If you feel that the spirit of WP:N does not match what's written in that article, you should take your complaint to its talk page. In my opinion, the sentence "Notability does not directly affect the content of articles, but only their existence" was given pride of place for a reason.

With respect to your third point:

Since when is engaging in a civil discussion, even a long one, an indication of misconduct? Aren't discussions like this supposed to make Wikipedia better? Your assumption that I've engaged with you here in bad faith is rather cynical. I'm actually enjoying this, and have learned quite a bit about Wikipedia and its policies.
If you follow the link I provided previously ([5]), you'll see that in the revision I want to revert to, the generative fixation hypothesis does not have its own section, and is mentioned in the Criticism of the Building Block Hypothesis subsection.
I misstated my objection to your sentence "This was added as a COI edit; linking to one's own PhD thesis almost never acceptable!". What I should have said is that an edit that violates WP:COI, while fraught, is not in and of itself bad. While the WP:COI article discourages conflict of interest editing, it does not forbid such edits, even at the article level: "Conflict of interest is not a reason to delete an article, though other problems with the article arising from a conflict of interest may be valid criteria for deletion." Wikipedia:COI#How_to_handle_conflicts_of_interest. This same section also has the following to say to editors dealing with conflict of interest edits: "Assuming good faith, start from the idea that the contributor was genuinely trying to help increase Wikipedia's coverage".

As far as I can see, the only rationale for barring me from mentioning my work is if you know that an impartial domain expert writing from a neutral point of view would not mention my work. Note: the domain expert is under no obligation to mention only work that has been cited. Since, by your own admission, you are not a domain expert, you aren't in a position to make this call.

So, I ask again, I'd like to roll the section on the building block hypothesis back to its state in the following revision [6]. Do you object? Keki Burjorjee (talk) 08:39, 18 June 2010 (UTC)

It's not about "having it both ways". "Notability" is not only a Wikipedia policy, but also a guideline that governs decisions about what's appropriate article content (along with common-sense, etc.). You're absolutely right when you state further down that "an impartial domain expert ... would not mention my work", at least in the context of this Wikipedia article, and at least in the context of today. That is because all content must be justified, and I'll say it again, there's no reasonable justification for mentioning a theory with no credence or acceptance in the field. Your theory is not even in the position where, for instance, it's been studied and rejected as invalid (even then it might be worth mentioning), it simply hasn't gained any exposure at all. Per WP:UNDUE and so on, Wikipedia should reflect this level of external coverage, i.e. none.

With respect to COI, please note that I'm not accusing you of bad faith; I'm suggesting that you're giving your own work undue importance. Why does your PhD merit mention in Wikipedia above the all the other hundreds or thousands that get written every year on the subject of genetic algorithms? If all we have to go on is your opinion that "this is the only plausible explanation for the capacity of genetic algorithms" (to paraphrase), then that's not good enough I'm afraid. That's where coverage by independent reliable sources would have helped.

So in summary, yes I would strongly object to rolling back, and I'd wager that if you started a request for comment, you'd get the same response from anyone else. In one sense, I suggest you do file a RfC if you want to carry this forward, because you aren't going to persuade me, and I guess you aren't going to persaude Erik either (not sure if he's following this or not). Oli Filth^{(talk|contribs)} 09:03, 18 June 2010 (UTC)

Oli, your latest comments suggest that you are not just, as you put it, nowhere close to being an expert on genetic algorithms, you actually have little, if any, knowledge about the field. If you read my edits (and what I've repeated at different points in our conversation) you'll see that I've never said that the "generative fixation hypothesis is the only plausible explanation for the capacity of genetic algorithms". What I've said is that the generative fixation hypothesis [7] is, *to date*, the only plausible explanation for the adaptive capacity of genetic algorithms *with uniform crossover*. Do you understand the difference between the general class of genetic algorithms and the specific class of genetic algorithms with uniform crossover (UGAs)? Do you know that UGAs frequently outperform genetic algorithms with one point crossover, even though according to the building block hypothesis they shouldn't? Are you aware that since 1989, when the utility of uniform crossover was first highlighted, there has been no serious attempt to provide an explanation for the adaptive capacity of GAs that use this form of crossover, much less explain why they outperform GAs with one-point crossover? Do you know how deeply the Building Block Hypothesis, despite its shortcomings, continues to influence contemporary research in evolutionary computation? Do you understand the repercussions of any new theory that lays the building block hypothesis to rest? Can you appreciate why such a theory might not get cited immediately? Do you care to educate yourself about "details" such as these?

You seem to care more about justifying your use of one-size-fits-all instruments to mechanically decide whether an edit stays or goes---no matter what the article. Having settled on the existence of citations as one of your tools, you wield it in a heavy handed way, feeling little obligation to dig further. Are you capable of recognizing that there are limits to the kind of policing that one can constructively do with little to no domain knowledge? Keki Burjorjee (talk) 04:08, 19 June 2010 (UTC)

I did some googling. It turns out that papers of mine that became chapters 2 and 3 of my dissertation have been cited in two recent Ph.D. dissertations: [8], and [9]. Does this satisify you? Keki Burjorjee (talk) 06:32, 19 June 2010 (UTC)

You're right, I know next to nothing about genetic algorithms. But actually, that hardly matters. What matters is that I know how to edit Wikipedia articles. One does not need to be well-versed in the field to perform standard copyediting. Call it "heavy-handed" if you like, but requiring stuff to be independently covered in non-trivial detail (etc.) is one of the most powerful tools for preventing articles from filling up with unimportant details, self-promotion, hoaxes, etc.

The specifics of what your thesis are totally irrelevant from the point of view of writing an encyclopaedia article. I've already covered what is relevant. Arguments such as "why such a theory might not get cited immediately?" are also irrelevant; either something is notable and covered now, or it isn't. The latter does not preclude it from becoming notable and covered in the future; if it does, then that's when it will be time to think about discussing it in a Wiki article. Oli Filth^{(talk|contribs)} 11:02, 19 June 2010 (UTC)

I reject your application of WP:N to article content, as does WP:N itself. Explicitly. At the top of the page. I have trouble understanding exactly which part of "Notability does not directly affect the content of articles, but only their existence" you don't understand. Wikipedia is not Citationpedia!

My goal is to improve Wikipedia for its readers, not to make it easy for you to do your chosen job. I wish you would recognize that there will be times when, as a non-expert, you won't be able to tell whether an edit benefits or hurts Wikipedia's readers. When this happens, it's okay to let others with more domain knowledge decide.

A question to ask yourself is what would have happened if Wikipedia had existed in the nineteenth century, and if in 1866, right after publishing his work on inheritance in pea plants, Gregor Mendel had made an edit about this work. Not as a separate article (that would violate WP:N), but within the main article on Inheritance. Something small, right after the section on blending inheritance and the troubles this (then dominant) school of thought was having explaining the existence of diversity in biological populations. Something like "Recently, a new hypothesis has been proposed that can explain the existence of diversity in biological populations. This theory is based on multi-generational experiments with pea plants. It holds that inheritance is particulate ....". Would such an edit have benefited or hurt Wikipedia readers? What happened, of course, is that Mendel's work went relatively unnoticed for 34 years, receiving only three citations during that time---all of them critical. It was nearly another two decades before Fisher, Wright and Haldane began weaving Mendel's work together with Darwin and Wallace's to give us the beginnings of the Modern synthesis. The Academy can be, and often is, stodgy. Wikipedia is a new frontier, in that it allows anyone to contribute relevant information that's in a reliable source. Please allow it to stay that way. If you don't dispute the relevance of my edits, and have no reason other than WP:COI to doubt my objectivity, kindly drop your opposition. Keki Burjorjee (talk) 05:46, 20 June 2010 (UTC)

Please don't trot out that analogy! New editors use it all the time, and the answer is always the same. It's very simple; Wikipedia doesn't give coverage to things that aren't yet covered by independent sources in non-trivial detail.

It's clear you have a problem with the applicability of WP:N. Perhaps you should re-read WP:UNDUE, and also WP:Fringe theories#Notability versus acceptance, which are essentially the application of WP:N to article content. They're the very antithesis of your notion that "Wikipedia is a new frontier", as the aim is for it to be the exact opposite. Oli Filth^{(talk|contribs)} 09:51, 20 June 2010 (UTC)

I'm interested in learning if others in the Wikipedia community share your ideas about what Wikipedia is and is not. And, more to the point, if they agree or disagree with your judgement in this specific case. I'm going to submit an RfC. My question is the same as the one on the talk page of the GA article: Should the following statement be removed from the end of the "Criticism of the building block hypothesis" section? "Recently, a new explanation---the Generative Fixation Hypothesis---has been proposed. [*]". (I agree that anything more goes against the spirit of WP:UNDUE)

I certainly agree about Wikipedia not being a "new frontier" in this sense - the mendel example is good: his theories, as important as they were ultimately, wouldn't receive much treatment on wikipedia until they had achieved a certain level of notoriety. wp:FRINGE and WP:SELFCITING really seem more applicable here than wp:UNDUE, although I suppose all are relevant. (Not to denigrate your work with a label like "fringe," KB!) I think wp:FRINGE makes it pretty clear that a reliable source is good enough for inclusion, and completed dissertations are wp:RS. Personally, per wp:COI, I would wait until someone else did the including, although this is easy for me to say given the likelihood of me ever writing papers such as this - anyway not a hard requirement. I have no opinion about the case at hand, regrettably I know very little about the theory behind GAs. Heh - my main complaint was the article forking, so I am happy since [10]. Thank you both for your work on this page. ErikHaugen (talk) 06:30, 21 June 2010 (UTC)

Thanks for your comments Erik. With respect to the "new frontier" bit, my words were used out of context. My full sentence was "Wikipedia is a new frontier, in that it allows anyone to contribute relevant information that's in a reliable source". By this I mean that a single domain expert is unlikely to be aware of relevant information published in all the reliable sources of her field. An encyclopedic entry that she submits is therefore bound to be lacking in some way. Wikipedia allows anyone to fill the gaps as long as he or she uses information that's in a reliable source, and writes from a neutral point of view. It is in this respect that it is a new frontier. Oli, FYI, I've added the following revised "keep" argument to the talk page.

The generative fixation hypothesis (GFH) is one of only two published, peer reviewed, non-folk explanations for the adaptive capacity of simple genetic algorithms (the other being the building block hypothesis).
The generative fixation hypothesis is the only theory to date that claims to account for the adaptive capacity of simple genetic algorithms with uniform crossover (UGAs).
In practice, UGAs frequently outperform genetic algorithms with one point crossover, even though according to the building block hypothesis, they shouldn't. The GFH explains why. In doing so, the latter hypothesis addresses an anomaly that has been outstanding for over twenty years.
Given the above, a domain expert writing from a neutral point of view about genetic algorithms would be remiss if she did not mention the existence of the GFH; unless, of course, she finds a glaring fault with the GFH that was missed during peer review.
By his own admission, Oli Filth knows next to nothing about genetic algorithms. He is, therefore, not in a position to determine if a domain expert writing from a neutral point of view would or would not make the statement in question.

Comments

Well, I believe the COI aspect of the situation becomes irrelevant if the topic has reliable and verifiable sources. If the two peer-reviewed sources seem verifiable, which peer-reviewed articles usually are, then it is worth a mention in the article. However, it should not be in the criticism section of another genetic method. It should have its own section like the block hypothesis does. — Parent5446 ☯ ^{(msg email)} 00:37, 22 June 2010 (UTC)

At this time, GFH should not be in the article. I have problems on several fronts.

First, the BBH is poorly covered although it is a long-standing theory. In addition, the criticism of the BBH does not appear neutral. Coverage and tone of BBH needs to be improved before new alternatives make sense. Coverage of the BBH does have merit, but a domain expert would not be remiss for failing to mention the existence of a new alternative theory. Second, the proponent of the GFH has a COI and does not come across as without passion or prejudice. Self promotion is a significant issue. Previous GFH edits also lack a clear explanation of what is going on, so the proponent lacks an objective perspective. Third, mere mention of GFH sounds more in compromise/appeasement rather than reasonable action.

Particularly disturbing is the proponent's ad hominem attack -- especially given the length of and the concessions in the user page discussion.

When the BBH coverage has improved, then mention of GFH could be appropriate. That coverage, however, should come from editors who have detachment and perspective (possibly gained from the passage of time). The significance of BBH (and GFH) in the article is not the promotion of hypotheses but rather an explanation of why genetic algorithms succeed.Glrx (talk) 19:32, 24 June 2010 (UTC)

PS. The proponent is also responsible for the non-neutral tone of the BBH criticism. See, for example, [Old Edit] Glrx (talk) 22:56, 24 June 2010 (UTC)

Conflict of interest is not an excuse to avoid neutral point of view. The very point of WP:COI is to protect WP:NPOV against editors who are vested in an opinion that gives them a biased point-of-view of the subject. However, just because an editor has a conflict of interest does not mean NPOV should be thrown out the window and we should just remove everything the COI author has put in, whether it has merit or not. GFH might be an actual legitimate alternative to BBH, and should be included in the article regardless of whether an author with a COI edited it first. What if somebody came in and had a COI with BBH, does that mean we should remove BBH as well? Of course not. Furthermore, the BBH coverage not being cleaned up is not an excuse to stop work on another section. In fact, I think the GFH section should take priority as you want to attain an NPOV before trying to push the article even further. I also have one more comment. I am not an established editor, so I cannot comment on the legitimacy of either techniques with confidence. But I think the primary thing that needs to be figured out is if the sources in this discussion, from both BBH and GFH, meet verifiability and notability standards. I know this has been said before, but I feel almost everybody here has been using the COI has a major point of argument, and meanwhile the COI is completely separate and unrelated. COI is a problem with an editor. NPOV is a problem with content. — Parent5446 ☯ ^{(msg email)} 02:54, 25 June 2010 (UTC)

COI and NPOV are not completely separate and unrelated. My comments were that the proponent has a COI. Furthermore, the proponent's content does not have a NPOV. Proponent is only one pushing GFH. Glrx (talk) 17:17, 26 June 2010 (UTC)

Another problem with rejecting edits primarily because of a conflict of interest is that it penalizes editors who in the interest of keeping things above-board write under their real names. It may have been possible to duck this entire controversy by making my edits under a pseudonym (the only reason that Glrx knows that I wrote the "Criticism of the building block hypothesis" section is because I wrote it under my real name. For the record, I also wrote most of the section on the building block hypothesis). My aim, however, has never been to get stuff into Wikipedia under the radar. I knew full well that using my real name would invite scrutiny. I used my real name anyway precisely because achieving a neutral point of view is important to me. Those who disagree with the substance of my edits are welcome to challenge any specific point, and/or make edits of their own. That they have not done so is no reason to keep a viable alternative to the building block hypothesis under wraps. Keki Burjorjee (talk) 04:23, 26 June 2010 (UTC)

Keki Burjorjee, I think you may be missing the point of WP:COI. As the author of the source being put into the article, you have a conflict of interest. Period. Because of this fact and this fact alone, policy recommends you not edit content relating to that conflict of interest. And I would have to agree with policy on this one. A conflict of interest can actually alter the tone of how you write and edit and thus make it POV, and in most cases this happens subconsciously without you even noticing. The point I was trying to make with the COI is that just because your edits on this topic should be rejected (read the COI policy for more details on that), that does not mean the content itself should be ignored completely. It means an editor w/o a COI should come in and make sure everything is kosher as far as NPOV goes. — Parent5446 ☯ ^{(msg email)} 16:18, 26 June 2010 (UTC)

The psuedonum/real name issue is tangential; it does not control COI or NPOV. Firth addressed edit frequency on the user page, so that is a rehash. Glrx (talk) 17:17, 26 June 2010 (UTC)

Switching to WP:N, proponent has not offered secondary source citations to his work. He has offered two primary source citations, but PhD dissertations are required to give short reviews of the literature. One citation was a mention in passing. A second citation recapped proponent's views, but kept its distance. Its tone did not sound in endorsement. The current situation appears to be simple: proponent wants to promote his original research. At this time, that effort is inappropriate. Glrx (talk) 17:17, 26 June 2010 (UTC)

Parent5446, I'm not disputing that I have a conflict of interest. I'm pointing out that rejecting edits primarily because of a conflict of interest is counterproductive because one can easily hide the conflict of interest by writing under a pseudonym. Rejecting edits primarily because of a conflict of interest might make sense in an online encyclopedia like Citizendium, which requires editors to use their real names. On Wikipedia, such a policy unfairly penalizes editors with conflicts of interest who, in the interest of transparency, make their edits under their real names. Regarding neutrality, I invite scrutiny of the substance of my edits from those with domain knowledge. I know I'm not infallible, and am happy to work with others to address any specific problems they see. Ideally, the article on Genetic Algorithms should reflect what a neutral domain-expert would write after digesting everything written in every pertinent reliable source. As I've mentioned previously, the domain-expert is under no obligation to mention only work that's been cited. The enforcement of such an obligation would be detrimental for readers. This is why, I believe, the sentence "Notability does not directly affect the content of articles, but only their existence" appears in the nutshell box at the top of WP:N. This sentence makes it clear that Wikipedia is not "Citationpedia". I think it's important to keep it from becoming so.

Glrx, please be careful about your use of phrases like "original research", which have a specific meaning within the current context. By virtue of being published in a reliable source, the GFH is not original research. See WP:No_original_research. Also, you you accuse me of making an "ad hominem" attack against Oli. Can you point to a specific instance where I maligned his character just to score a point? Noting that Oli does not have domain-knowledge, or that he is approaching a subtle issue in a heavy-handed way is not an ad hominem attack within the context of this discussion. See Ad hominem. You, also talk about the long length of the discussion on Oli's talk page, and mention "concessions made". As I've previously said, the length of a discussion is not an indication of any kind of wrongdoing. And, as far as I can tell, I'm the only one that made any concessions in that discussion. Finally, I don't see how the number of edits I make is relevant. Are you saying that in one sitting I should be making one big edit as as opposed to several incremental ones? Keki Burjorjee (talk) 20:32, 26 June 2010 (UTC)

Well, I'll first address Keki Burjorjee: I somewhat disagree. You are correct to say that rejecting edits primarily because of a COI is counter-productive to the article, and that it penalizes authors who edit under their real name, but it is necessary in order to maintain POV. The editor, real name or not, should not be editing an article he or she has a COI with, so the penalization of editors under their real name is justified. We could go on and argue how to determine whether an editor editing under a pseudonym has a COI or not, but that is another argument completely. I do agree with you on the WP:N point though. However, what we still need to determine is if the GFH sources are WP:V.

In response to Glrx: that's exactly what I was trying to say. So now if the proponent has a COI, but the content he is contributing does not violate NPOV, then there is no reason to reject his edits, specifically because the point of the COI policy is just to protect NPOV. And as far as secondary sources go, that is not WP:N but rather WP:V. Notability is not in question here, verifiability is. And you might want to specifically look at WP:SECONDARY. Primary sources are just as valid and correct as secondary sources are as long as the editor does not attempt to analyze or interpret the primary sources, which would be WP:OR. — Parent5446 ☯ ^{(msg email)} 21:44, 26 June 2010 (UTC)

I'd like to pick this thread back up. Parent5446, I'm not sure I understand what you mean when you say "However, what we still need to determine is if the GFH sources are WP:V". WP:V says "The threshold for inclusion in Wikipedia is verifiability, not truth—-whether readers can check that material added to Wikipedia has already been published by a reliable source, not whether editors think it is true". By virtue of being published in a Ph.D. dissertation deposited with UMI (see here), the GFH has been published in a verifiable source; so, I don't quite understand your concern about verifiability. —Preceding unsigned comment added by Keki Burjorjee (talk • contribs) 05:19, 1 July 2010 (UTC)

Oh, no, I totally agree with you. My reply was a little ambiguous. If you read the second paragraph, my response to Glrx, you see what I mean. It's just I wanted to respond to the both of you separately. This is exactly why I am arguing for the inclusion of the material. — Parent5446 ☯ ^{(msg email)} 00:28, 4 July 2010 (UTC)

Thanks for this clarification. Keki Burjorjee (talk) 18:29, 4 July 2010 (UTC)

Consensus?

My readinging of the discussion is the consensus is no, the GFH should not be mentioned in the article at this time. Some commenters believe GFH does not yet satisfy WP:N. One of those commenters believes that GFH should not be a separate article, but takes a neutral position about whether GFH should be mentioned in this article. Two commenters believe that in this case WP:N should be a consideration should preclude mention at this time. Two commenters (proponent being one) believe that WP:N is not an issue. One of those commenters believes WP:V is the issue and is unsure that it is satisfied. Most commenters are concerned about WP:COI even though they recognize it is not a reason for an outright rejection. COI often compromises WP:NPOV, and that appears to be an issue here. These reservations would vanish if there were independent, reliable, secondary sources that supported the views and findings of GFH. Glrx (talk) 19:24, 3 July 2010 (UTC)

No consensus has been established yet. From what I can see there are only three editors in this entire argument, Keki, myself, and you. The first two agree that WP:N is not an issue and that the sources meet WP:V, while you disagree on that issue. I might have missed some editors if they were in a different section of discussion. Furthermore, while you are right that WP:COI is not a reason for outright rejection and that COI often compromises WP:NPOV, we do not need "independent, reliable secondary sources". From WP:V, information needs a "reliable published source", but we already have that. — Parent5446 ☯ ^{(msg email)} 00:34, 4 July 2010 (UTC)

I interpreted your statement, "However, what we still need to determine is if the GFH sources are WP:V.", to be that you were unsure about WP:N. Other editors have commented on Oli Filth's talk page. Glrx (talk) 05:15, 4 July 2010 (UTC)

Did you read the paragraph that came after that statement? (As well as my reply to Keki when he pretty much said the same thing you just said.) — Parent5446 ☯ ^{(msg email)} 05:26, 4 July 2010 (UTC)

Yes. Sorry, but my ref to WP:N was a typo that should have been WP:V. The topics you reference were previously addressed on OF's page. Glrx (talk) 17:38, 4 July 2010 (UTC)

Oh, got it. I hate it when I make typos. — Parent5446 ☯ ^{(msg email)} 03:09, 5 July 2010 (UTC)

Glrx, I'm still trying to understand why you believe my edits violate WP:NPOV. Is this just a vague feeling you have, or are you able to point to some sentence, or set of sentences, and tell us specifically how they violate WP:NPOV? What specific details would you like to see in their place? Keki Burjorjee (talk) 06:26, 6 July 2010 (UTC)

Glrx, it's been a little over a week since I posed the questions above. During this time you have not provided specific details about how the proposed edit (which is verifiable), or my past edits (which are also verifiable), violate WP:NPOV.

I think we can agree that the ultimate question is "Does the inclusion of these edits improve or damage Wikipedia?" (see Wikipedia:Ignore_all_rules). I hold that their inclusion improves Wikipedia, and that the information provided is more helpful than much else that is currently in the Genetic Algorithms article. Unless you or Oli can provide substantive evidence to the contrary, the edits should stand. Keki Burjorjee (talk) 17:12, 14 July 2010 (UTC)

include - the notability is not the issue here. It is a minor field and verifiability is more appropriate. "Material such as an article or research paper that has been vetted by the scholarly community is regarded as reliable." "Finished Ph.D. dissertations, which are publicly available, are considered publications by scholars and are routinely cited in footnotes. They have been vetted by the scholarly community; most are available via interlibrary loan. UMI has published two million dissertations since 1940." The people that are included on the dissertation committee are pretty notable though: Jordan Pollack[11], Harry Mairson [12], Timothy J Hickey[13](wikipage up for deletion), Lee Altenburg[14]Chaosdruid (talk) 01:35, 20 July 2010 (UTC)

The point here is that COI is a policy and has some basic rules. First of all any editor that has a specific interest in any article should declare so on the talk page. Changing your name would not evade that rule. You would still need to declare your interest in the article you were editing and changing your name to avoid declaring COI, avoiding declaring COI or editing under two names would indeeed be problematic. The first two would be seen as bad faith and the third would be puppetry. COI in itself does not prohibit editing on an article in your specialist field. If it was an article on you, or about your company that you owned/were director of or about your own theory, that would be COI. I would also ask if there are any editors who have already edited the article who have not declared any COI themselves - are there any editors who are actually involved in the field of Genetic algortihms that have not declared a possible COI from working with GA throught the BBH school? We often ask for an expert opinion on a subject/article and who better to ask/edit than an expert? well as you will see that is not always such a good thing...

The real problem, as stated before by others, is POV. A declaration of COI, especially in a field of research/theory where the normal editor would not necessarily be learned, would allow others to understand that your edits were perhaps POV. The problem here is that the edit you want included , when taken on its own, appears as COI. In my opinion that is not the case - it is a point of view that is backed up by being a published papaer. As such it will have been (I hope) scrutinised by peers and more learned reviewers and should have passed muster as being factually correct and based upon fact.

Imagine a situation where we have five astrophysicists in a room. The question id given to them:-

"What theory best dscribes the nature and matter of the universe? : Write it here on the board"

You can imagine the mayhem as the string theorist, bubble universe, multiverse and manyworlds guys all debate heatedly for weeks without actually agreeing on anything, the constant bickering over "you haven't got any proof" & "not supported by the evidence", the writings crossings out and arguing over who has the pen all of which may even lead to fisticuffs lol.

That is the problem. It is the point of view. If the question had been "What are the theories of the makeup of the universe? : please write them here and quote evidence for each." then the POV becomes less important. Now it is verifiability. Anything written on the board that cannot be sourced from a notable place would be stricken off. If you do not declare your COI (as say a multiversist) and lets say start to delete the string theory entries : once your COI is discovered it will be treated in a very different was than someone who just "read somewhere that string theory was out of date". We are not saying that your view is necessarily biassed, just that your views should be more strictly scrutinised to ensure they are not POV and that can be a difficult matter if someone is an expert in a field.

I do not believe that the addition of the material is a problem as long as it states that it is just a current theory in development but this should not be used as it would perhaps cause a precedent where every single theory of smoething which has been published as a paper can be used in an article. You can imagine how many would be suddenly inserted into the Universe page...

I would say, however, that edits to the article have made it almost impossible for the layman to understand, a common mistake for specialists, and suggest addressing these issues:-

"In a genetic algorithm, a population of strings (called chromosomes or the genotype of the genome), which encode candidate solutions (called individuals, creatures, or phenotypes) to an optimization problem, evolves toward better solutions" : This needs explaining more, what is a string, what is a candidate solution - maybe give an example of such an optimisation problem (and one that is easily understood)
Edits have removed the links to heuristics which need to be explained or relinked.
Methodology - the opening sentence has commas all over the place and really is a little too long and this problem runs through the article
Peacocking - any use of words which increase puffery as per WP:PUFFERY and WP:PEACOCK

It is apararent to me that fitness is the problem with Genetic algorithms. In reality selsction is not always based on fitness. Lets take two negative examples. A fit organism exercises regularly, eats well and endeavours to avoid unhealthy things. Unfortunately for the white moth, industrial pollution plays a hand and the brown moth is better hidden on the dirt surfaces. The birds feding on the moths find the white moth more easily and eat them all. This is not dependent on fitness. A standard distribution curve is probably not reallyy relevent here but mutation is often the way forwards. Without mutation adaptability is not possible. Mutations can be both at the left or right of a standard distribution curve for a population, where non mutated "fitness" is in the middle, and without taking into account these factors it is not surprising that a procedure which only measures successive "better fitness" does not give expected ditributions of results. Chaosdruid (talk) 01:35, 20 July 2010 (UTC)

Consensus (take 2)

I count three in favor of the inclusion of the edit (Parent5446, Chaosdruid, and myself), one who is ambivalent about it (Erik Haugen), one who is actively opposed (Glrx), and one whose continued opposition is in question because of his silence in this discussion (Oli Filth). The majority opinion at this point seems to be that the edit should be included. Keki Burjorjee (talk) 16:40, 27 July 2010 (UTC)

Glrx has reverted the edit on the grounds that neither I nor Oli Filth should declare a consensus, presumably because this discussion originated in a dispute between Oli Filth and me. I can find no Wikipedia guideline that supports Glrx's position. Nevertheless, if someone else can discern a consensus to this discussion, please describe it. Consensus is not unanimity seems to apply here.

Glrx, I remain open to criticism of the substance of my edits. Can you provide material evidence of non-neutrality, or evidence that learning about the existence of the generative fixation hypothesis is detrimental to Wikipedia's readers? I think we can agree that reverting the edit while remaining silent on this thread is not constructive (please see WP:DRNC). Keki Burjorjee (talk) 05:10, 28 July 2010 (UTC)

See WP:SCOIC stating that "Once you have presented your case, it may be best to take a back seat and minimise your participation in the discussion, to avoid any perception or accusation of undue advocacy or pushing an agenda." The RfC asked editors to comment on this issue. It should not be an open invitation for you to engage in arguments with those editors. There are many watchers of the GA page; very few have engaged in this debate. When I thought there was a consensus, I posted that on the talk page to test the waters. There was disagreement. Fine. Glrx (talk) 17:51, 28 July 2010 (UTC)

You are continually engaging in narrowly focused arguments. You are demanding that other editors prove some point or else concede your edits. You want me to cite examples of your COI even though you stated, "I concede that my most recent edits may (kinda sorta) give the impression that the GFH is "equal" to the BBH in terms of adoption." I'm looking more at the big picture. What secondary sources have covered the GFH? If the BBH and GFH are the only contenders in town, then a secondary source should be easy to find. Where is it?

WP:UNDUE teaches

Undue weight applies to more than just viewpoints. An article should not give undue weight to any aspects of the subject but should strive to treat each aspect with a weight appropriate to its significance to the subject. For example, discussion of isolated events, criticisms, or news reports about a subject may be verifiable and neutral, but still be disproportionate to their overall significance to the article topic. This is a concern especially in relation to recent events that may be in the news. Note that undue weight can be given in several ways, including, but not limited to, depth of detail, quantity of text, prominence of placement, and juxtaposition of statements.

From Jimbo Wales, paraphrased from this post from September 2003 on the WikiEN-l mailing list:

If a viewpoint is in the majority, then it should be easy to substantiate it with reference to commonly accepted reference texts;
If a viewpoint is held by a significant minority, then it should be easy to name prominent adherents;
If a viewpoint is held by an extremely small (or vastly limited) minority, it does not belong in Wikipedia regardless of whether it is true or not and regardless of whether you can prove it or not, except perhaps in some ancillary article.

Keep in mind that, in determining proper weight, we consider a viewpoint's prevalence in reliable sources, not its prevalence among Wikipedia editors or the general public.

Where is the prevalence and acceptance of GFH in reliable sources? I've commented about the two PhD citations already. Where are the prominent adherents? (I disagree with the notion that members of a PhD committee are adherents.) Where are the writings that have reviewed GFH and agreed with it. To turn things around with your logic, if you cannot prove prominent adherents, then the GFH viewpoint is held by a limited minority and should not be in Wikipedia.

Glrx (talk) 18:09, 28 July 2010 (UTC)

The statement that OF's continued opposition is in question due to his silence is outrageous. OF has already engaged in a lengthy and complete discussion. Glrx (talk) 18:18, 28 July 2010 (UTC)

This is getting a little out of hand now. It takes two to argue ! The amount of comments you have added, plus that huge chunk of text from OL page are not helping us achieve consensus nor is posting out of chronological order helping editors to follow the discussion.

Are we at least agreed that the PhD paper is a reliable source according to WP:MoS as per my note earlier

""Material such as an article or research paper that has been vetted by the scholarly community is regarded as reliable." "Finished Ph.D. dissertations, which are publicly available, are considered publications by scholars and are routinely cited in footnotes. They have been vetted by the scholarly community; most are available via interlibrary loan. UMI has published two million dissertations since 1940.

If so then at least we can proceed in narrowing down what the consensus is on wording of inclusion as there is a massive difference between portraying it as a major competitor and mentioning it in passing.

Can I also point out that any COI should be declared - especially from any editors involved in this discussion.

I have also responded on your talk page Glrx

Chaosdruid (talk) 18:54, 28 July 2010 (UTC)

Glrx, the sentence you quoted from WP:SCOIC ("Once you have presented your case, it may be best to take a back seat and minimise your participation in the discussion, to avoid any perception or accusation of undue advocacy or pushing an agenda.") is advice to disputants to help them avoid precisely the situation that seems to be arising here --- where an editor takes any challenge as prima facie evidence of a disputant's non-neutrality, hardens his or her opposition, and obstructs further discussion. Your contention that the quoted sentence requires me to be silent in this discussion is a regrettable case of the tail wagging the dog. Additionally, please look at the sentence that follows the one you quoted. It says, "Of course, if there are indications of any misunderstanding or misconception, it is reasonable to politely address this". Unfortunately, your comments have given me cause to set matters straight on multiple occasions (present one included), and seek clarification from you on others. Your claim that my efforts toward this end are scaring away other potential discussants is extremely speculative, to say the least.

In response to your statement You are continually engaging in narrowly focused arguments. You are demanding that other editors prove some point or else concede your edits. You want me to cite examples of your COI even though you stated, "I concede that my most recent edits may (kinda sorta) give the impression that the GFH is "equal" to the BBH in terms of adoption.": I believe it is you who has a narrow focus. I'll get to that in a bit. First I'd like to take up the matter of my neutrality (clearly, neutrality, not COI, is the issue here. I don't dispute that I have a COI). Please note that the edits referenced in the sentence you quoted were ones in which I gave the GFH its own section. The referenced edits do not include the edit that is the subject of this discussion. In retrospect, it was wrong to give the GFH its own section, because doing so at this point in time has the effect of giving this hypothesis undue weight. As the quote above shows, I conceded as much in my discussion with Oli Filth. And if you look at the text of the discussion, you'll see that I did so readily. You parade this concession as evidence that I cannot be objective about my own work. I happen to think it constitutes evidence that I can.

If you think my other edits are not neutral, it is up to you to tell us specifically why, and to explain what you would like to see in their place. You may indeed have a valid point, and I'd be happy to work with you to achieve neutrality (believe it or not, drawing you into an endless argument is not my goal). This process, not knee-jerk deletion, would be in keeping with the spirit of Wikipedia.

Onto your narrow view about what does and does not belong in Wikipedia. The ultimate question, and I keep coming back to this, is whether or not an edit is helpful to Wikipedia's readers (See WP:Ignore_all_rules). In some cases it will be possible to answer this question without knowing anything about a field. In other cases (present one included) domain knowledge is required. You've given no indication that you have any. An ability to count citations is no substitute for actual knowledge about a field and its history. Again, Wikipedia is not Citationpedia. Members of the genetic algorithms community are fully capable of taking me to task for unhelpful edits. They are especially likely to do so if the edits are about a matter as important as this one. Their silence during the months when my edits were visible speaks for itself (See WP:Silence_and_consensus).

Perhaps I am wrong in my hunch that you have no domain knowledge. If so, please tell me which of the following statements you disagree with:

Since the early '90s when significant theoretical and experimental problems with the building block hypothesis were first noted, no alternative peer reviewed explanation for the adaptive capacity of genetic algorithms has been proposed; that is, until now.
Aside from the GFH there is no peer reviewed hypothesis that professes to explain the adaptive capacity of the genetic algorithm with uniform crossover.
It's not like new peer reviewed explanations for the adaptive capacity of genetic algorithms get published everyday. When one does get published, readers would probably find it helpful to learn of its existence.

If you continue to oppose my edit, but don't know enough about the field to dispute any of the above statements, then I'd have to say that you're putting other concerns ahead of the best interests of Wikipedia's readers; in other words, that you are, once again, letting the tail wag the dog.

Finally, about Oli Filth's silence: His continued silence even after his opposition has been called into question by Parent5446 and me is hard to argue with. It's his responsibility to speak up if he feels that Parent5446 and I have erred in our assessments. Keki Burjorjee (talk) 08:47, 3 August 2010 (UTC)

There is no question that the paper meets WP:RS - I think that we are also agreed that any mention should not be in a separate section or subsection. This leaves us with the size of addition to decide - 1. a separate paragraph or 2. a sentence added to an already existing paragraph.

If there is any weight to be added it should be option 1, a paragraph, and if there is little weight it should be option 2. My personal view is that at this moment in time it should probably be option 1 although no more than 40 words maximum and an external link or two in the links section added. If, over time, it is shown that the theory is accepted and expanded we can increase the size or even give it its own section.

At this point I think that KB should write his proposed addition here and then we can see if we agree on content, how big, and where it shold go.

Chaosdruid (talk) 21:11, 3 August 2010 (UTC)

Please read the thesis of this discussion. The proposed addition to the article ("the following statement") is clearly identified.

Where are the prominent adherents that show the GFH can avoid the "vastly limited minority" viewpoint label? Alternatively, where is the work reviewed in a secondary source? KB is arguing the truth and significance of his work, but that is not the issue in an encyclopedia.

I have no trouble with continued silence. I certainly do not see silence implying the reversal of a well-stated position.

Why is it that that KB questions the qualifications of OF and me ("don't know enough about the field"), but he does not question the qualifications of other editors? These issues are diversions that hide the main issue. Where is the evidence that GFH merits coverage? If the GFH is as significant as KB claims, then it should come to the attention of those compiling secondary sources.

Glrx (talk) 18:37, 9 August 2010 (UTC)

? I have read it, and watched the video. You seem to be getting a little confrontational though - I am a neutral editor.

The proposed addition (above) is not being debated. Based on the facts that the PhD meets WP:RS I cannot see a problem with including a sentence of 20-40 words that would mention the thesis on GFH. Surely it is whether mentioning the work falls under the purview of OR or not that is the issue and whether it may meet notability guidelines or not. I, as an uninvolved editor, would not fall under OR or SYN or COI if I mentioned the work as a related topic in passing. I do not see any difference to me making the sentence or us agreeing on one that KB wishes to put forwards.

As for the issue of "vastly limited minority" there is no harm in mentioning a minority viewpoint if it is backed by a reliable source is there ?

You ask "Have I read the thesis" - what point do you wish to make on that matter?.

I have no problem with a sentence like "Recently a theory has been put forwards..." and use the thesis as a ref. I realise you do have though and am trying to see if there is any way an agreement can be made.

Chaosdruid (talk) 19:15, 9 August 2010 (UTC)

Sorry, I used "thesis of this discussion" to refer to the question presented on this talk page rather than KB's thesis. The debate is about including a specific sentence; your proposal sought more words from KB -- which is more that KB sought.

I see harm with WP:UNDUE. Where is the evidence that shows the GFH qualifies as a significant minority viewpoint (and not a viewpoint held by an extremely small number)?

Glrx (talk) 20:03, 9 August 2010 (UTC)

Same here - it is not mentioned in any other sources other than by KB and mainly on KB's sites (University/blog/yahoogroup/ai etc)

It would be a slightly different matter if one of the panel that was on his thesis had mentioned it in a news item, published work/site etc or somewhere else but I did not really want to start putting ideas in peoples heads lol :¬) Chaosdruid (talk) 20:50, 9 August 2010 (UTC)

I don't think the matter would change if one of the panel members mentioned the work. Mere mention does not make an adherent. Furthermroe, it must be easy to name significant adherents (plural). (I'd also want to see significant independent adherents.) In fairness to KB, his work is cited in two PhD theses. Those citations, however, do not offer a review of KB's work but rather just point to and describe the work. (IIRC, they do not use the term GFH.)

I want to see either prominent adherents or independent survey articles about GA (or BBH) describe KB's work and comment positively on its significance. (I might even accept the GFH appearing in computer science course outlines.) Until that time, I believe Wikipedia should be silent about the GFH. Is that an unreasonable position? Glrx (talk) 21:33, 9 August 2010 (UTC)

All I can say is that there are quite a few "I want" in there lol

But seriously, independent adherents - One of two off the board mentioning it in article or summaries or webpages would be independent. It would not be right to assume COI just because they were on the board of the PhD ?

From my perspective I do not see the harm in a passing mention that a theory has been put forwards, or similar, as I do not believe it is against BBH but rather it refers only to the problem of the unexplained high positives in fitness function:

"In particular it is difficult to understand why these algorithms frequently succeed at generating solutions of high fitness when applied to practical problems"

I also believe that you are correct that to give it a full explanation at this point would be incorrect and that for full explanation it would need secondary sources of more high calibre.

Chaosdruid (talk) 22:14, 9 August 2010 (UTC)

I'd like to pick this thread back up. With respect to the Jimbo Wales quotation, Glrx, you seem to be trying to paint the generative fixation hypothesis with the same brush that might be used to paint any old point of view, even one concocted on a whim. The generative fixation hypothesis is not merely a "viewpoint". It is a scientific hypothesis vetted by a committee of scholars, and published in a reliable source. (Your blanket statement that Ph.D. theses do not receive the same scrutiny as journal articles is misguided. Surely you can appreciate that that the quality of scrutiny in each case is dependent on the people doing the scrutinizing). You have repeatedly asked for evidence of the notability of the generative fixation hypothesis. The Wikipedia guideline on notability, however, is quite clear that notability is a filter for article existence, not article content (see the nutshell box at the top of WP:N). The filters for article content are Verifiability, No original research and Neutral point of view; WP:V states, "Verifiability is one of Wikipedia's core content policies, along with No original research and Neutral point of view. Jointly, these determine the type and quality of material that is acceptable in articles". The sentence about the non-applicability of WP:N to article content in the nutshell box of WP:N indicates that the prevailing vision of what Wikipedia should be is more expansive than the one you currently subscribe to.

I'm not saying that all edits that pass WP:V, WP:NOR, and WP:NPOV merit inclusion; only that if you insist on pruning an article while remaining uneducated about the domain, then your activity should be limited by these three guidelines, not WP:N. Anything more is a lot like driving your friends home while drunk. Sure, you may think you're being helpful, but there is a high probability that you will cause your friends harm. To come back to the bottom line, if you continue to shoot down the proposed edit, you must provide one or more concrete, domain-specific reasons why learning about the existence of the generative fixation hypothesis is to the detriment of readers of Genetic algorithms. I can provide many reasons why it is beneficial (more beneficial, at any rate, than a lot else that's in the article). Keki Burjorjee (talk) 23:06, 22 August 2010 (UTC)

Glrx, has again reverted the proposed edit. Grlx, like it or not, your reverts make you a primary disputant in this case. Non-responsiveness on this thread followed by unexplained reverts goes squarely against the spirit of Wikipedia. You have rebuffed attempts by other editors to reach a consensus, including an attempt by an editor who reached out to you on your talk page (Chaosdruid). Furthermore, you seem to feel no obligation to remain engaged on this talk page. Continuing in this vein will get you reported to the arbitration committee. WP:Dispute_resolution says, "Talking to other parties is not a formality; it's an imperative to the smooth running of any community. Not discussing will make people less sympathetic to your position and may prevent you from effectively using later stages in dispute resolution."

I'll state my position again. The proposed edit accords with Wikipedia's three content policies: WP:V, WP:NOR, and WP:NPOV. Mentioning the existence of the generative fixation hypothesis in the article on Genetic_algorithms is, in my judgment, in the best interests of Wikipedia's readers, especially ones who are curious about the adaptive capacity of genetic algorithms with uniform crossover (which is currently unexplained). Do you have a good reason (not a speculative one) for believing otherwise? If so, please state it. Keki Burjorjee (talk) 02:06, 29 August 2010 (UTC)

The revert is not unexplained. You are repeating your position. You want to add material, so it is your burden to show merit and not mine to show there isn't. Although it has been requested, you have not provided a secondary source or other evidence that shows prominent adherents. You asked me (and others) to weigh in on your RfA. I am not a "primary disputant". I've weighed in that mention is premature. The matter has been discussed with Chaosdruid and others. Where is the consensus for your addition?

You keep coming back to arguing the truth of your work. That is not the issue. Glrx (talk) 02:45, 29 August 2010 (UTC)

Glrx, an RFC is indeed an invitation to, as you put it, "weigh in" on a dispute. By reverting the edit concerned, however, you are doing more than weighing in. Your reverts make you a primary disputant. You and I clearly seem to have a fundamental disagreement about where the burden of proof lies here. I haven't seen any Wikipedia policy that leads me to believe that I need to provide evidence that shows "prominent adherents" in order to merely mention the existence of the generative fixation hypothesis on the Genetic algorithms page; rather, my reading of Wikipedia's policies and guidelines leads me to conclude that that the satisfaction of WP:V, WP:NOR, and WP:NPOV is sufficient. It seems you disagree. In the interest of coming to a resolution would you be open to mediation (WP:RFM)? Keki Burjorjee (talk) 04:46, 29 August 2010 (UTC)

convenience break

Comment. I found this under the Mediation Cabal heading, but I don't want to take on the role of a formal mediator. But as an outsider my 15-30 minute first impression is:

The Ph.D. thesis is a valid source that should be taken seriously and can be referenced. Much of all research is done by Ph.D.s and a thesis represents a major statement of opinion in a field.
The sentence "Recently, a new explanation---the Generative Fixation Hypothesis---has been proposed." is far too pretentious. It takes the reader's time to name a theory and claim that it's a new explanation. The Ph.D. thesis does present it as an alternate explanation, but a certain amount of self-promotion tends to be tolerated in academia; I'd want to hear an outside source say that it is such a dramatic break. I should compare this to the other criticisms of the building block hypothesis given in the text, which are simply in the form that so-and-so says quote... without saying that they proposed a major alternative idea. I am also somewhat influenced by the existence of a Google Group for the Generative Hypothesis,[15] not much used, which supports my thought that some self-promotion is going on.
The thesis discusses "hyperclimbing" in some length. It's not entirely clear to me from a quick read whether this is presented as a Burjorjee invention or whether it was pioneered in the Fogel, 2000 reference. If other sources exist, the idea of hyperclimbing should be introduced separately.
A sentence about the thesis might say that the author "argues" that the hypothesis "can account for the adaptive capacity of a wider class of simple genetic algorithms [including] simple genetic algorithms that use uniform crossover", "presumes less about the distribution of fitness over the chromosome set [and does not] presume the existence of a hierarchy of building blocks", and "can successfully pass a demanding test of validity, one that involves the application of a simple genetic algorithm with and without a mechanism called clamping to large, random instances of MAXSAT." It probably should say that "In breaking from the building block hypothesis, the generative fixation hypothesis reverts back to two classic positions in population genetics: 1. That fixation is the vehicle by which adaptive gains are secured. 2. That the function of recombination is to prevent hitchhiking."

Bottom line: include the reference, but reconsider what text should come with it. The text may well be longer, but not focus so much on promoting the hypothesis. Wnt (talk) 07:03, 26 October 2010 (UTC)

Thanks for your comments Wnt. What do you think of the following statement? "A recent explanation[*] holds that genetic algorithms, including those with uniform crossover, work by implementing a heuristic called generative fixation extraordinarily efficiently. Unlike the building block hypothesis, this explanation does not presume the existence of building blocks and hierarchies thereof. In breaking from the building block hypothesis, the generative fixation hypothesis reverts to two classic positions in population genetics: 1) That fixation is the vehicle by which adaptive gains are secured. 2) That the function of recombination is to prevent hitchhiking. The major difference between the generative fixation hypothesis and the orthodox view in population genetics is that whereas the former holds that a small group of unlinked genes can be a Unit of selection, the latter discounts this possibility out of hand".

I believe this statement is more informative, and less "promotional sounding" than the one I previously proposed. Keki Burjorjee (talk) 00:26, 1 November 2010 (UTC)

Wow. That's a bit intimidating. ;) And I'm not really qualified to mess with it. Even so, you say "generative fixation" over and over again without saying very clearly what that means, and still emphasize the different hypotheses and the "breaking" between them. I would think to trim this down a bit:

"An alternate explanation dismisses the importance of building blocks and hierarchies thereof and proposing that a small group of unlinked genes can be a Unit of selection. [*] Hypothesizing that a heuristic called "generative fixation" is implemented with great efficiency, it returns to classic positions in population genetics: 1) That fixation is the vehicle by which adaptive gains are secured. 2) That the function of recombination is to prevent hitchhiking."

I'm not sure that's a good edit (I actually don't understand how a group of unlinked genes works as a unit of selection), nor that it is edited enough, but I hope it's a step in the right direction. Wnt (talk) 04:36, 1 November 2010 (UTC)

Unfortunately, the edit you propose doesn't work so well, Wnt. Taking a step back, I guess I don't understand your objection to my calling the generative fixation hypothesis by its name and providing a reference to my dissertation where this hypothesis is described in detail. Others have opined that describing this hypothesis in significant detail on Wikipedia at this point in time would give it undue weight, and I agree with them. I'm also puzzled by your opinion that I should provide more information about the GFH, but should avoid talking about the way it differs from other theories. IMO situating the GFH with respect to other relevant theories is an effective way to convey its gist. None of what I say in the passage above is Original Research (as defined in WP:NOR). All of it appears in my dissertation, which has been vetted. So, given that you hold that a Ph.D. dissertation is reliable source, I'm not sure I understand your objection.

In any case, what do you think of the just following sentence? "A recent explanation[*] holds that genetic algorithms, including those with uniform crossover, work by implementing a heuristic called generative fixation extraordinarily efficiently." I think it would be apparent to readers that if they want to know more about the heuristic they should look up the reference. Thanks for your involvement. Keki Burjorjee (talk) 03:24, 2 November 2010 (UTC)

It's very possible that my wording had its own problems. Certainly I see no problem with the last one-sentence version above, and I feel it is definitely better to include that sentence than to leave it out. Wnt (talk) 23:38, 5 November 2010 (UTC)

Thanks for this clarification Wnt. Of the active participants on this thread, I now count 4 in favor (Wnt, Chaosdruid, Parent5446, and Me), one neutral (ErikHaugen), and one against (Glrx). This discussion has been open for long enough and needs to be closed. I believe that a neutral party reading this thread would be hard pressed to argue that the WP:ROUGH_CONSENSUS is against the inclusion of the edit (see also WP:NOTUNANIMITY). Glrx, you and I clearly have a difference of opinion on when a scientific hypothesis merits being mentioned on Wikipedia. I believe that the satisfaction of WP:V, WP:NOR, and WP:NPOV are sufficient---especially in a domain that is starved for fresh ideas. Others on this thread share this view. I can find no support for your position that a WP:V scientific hypothesis cannot be mentioned in a Wikipedia article unless it has prominent adherents. Keki Burjorjee (talk) 22:17, 13 November 2010 (UTC)

Short and succint. Seems fine to me :¬) Chaosdruid (talk) 00:34, 14 November 2010 (UTC)

No problems here either. Hopefully we have finally ended this argument once and for all. — Parent5446 ☯ ^{(msg email)} 00:55, 14 November 2010 (UTC)

Once again, KB is trying to declare consensus. I have reverted. KB has also ignored the result of his trying to resolve a content dispute by bringing an etiquette alert. Glrx (talk) 01:21, 14 November 2010 (UTC)

Material from the alert:

Your dissertation is not an RS. There is no RS until the work is published in a reputable journal. Even then there is a high threshold for adding accounts of one's own work. In short, even if some other editors are naive enough to think that what you're doing is okay, I definitely don't. If you want further opinions about the usability of a dissertation as a source, you could ask at WP:RSN. Looie496 (talk) 23:19, 18 September 2010 (UTC)

According to WP:RS#Scholarship, a Ph.D. dissertation is a reliable source. Keki Burjorjee (talk) 23:59, 18 September 2010 (UTC)

Huh. Whoever wrote that doesn't know much about dissertations. Even the best of them usually contain some cruft; the worst ones that I have seen are pathetic. Anyway, why isn't there a journal paper about the work in question? Looie496 (talk) 00:08, 19 September 2010 (UTC)

I've had other priorities. Regarding your comment about dissertations: the same can be said to apply to journal articles; I can point you to some doozies in the field of Evolutionary Computation. At least with dissertations one knows who the reviewers are. My point is that when evaluating the peer-reviewed publications of a given field, it's hard to determine what's reliable and what's not without at least some domain knowledge. Wikipedia's baseline policy of considering Ph.D. dissertations reliable makes sense to me. Keki Burjorjee (talk) 00:36, 19 September 2010 (UTC)

Regardless of an individuals feelings the fact remains that a PhD dissertation is a reliable source. Not only that but the individuals sitting on the panel which have taken the dissertation are extremely notable in the field and in Artificial intelligence in general.

The matter should be weighted, verifiable and sourced correctly. The original statemenet included by KB was made and as a result GFH appeared as an alternative to the whole BBH subject matter. This was incorrect as the GFH is only a fix for the problem of "too many positivies" within BBH - this was remedied and the proposed material was reduced to a paragraph within the original section. I pointed out that a paragraph on GFH is probably too much at this point and proposed it should simply be a sentence tagged on the end of the BBH to state that it exists as a fix for the too many positives problem within BBH. This pretty much mirrors the proposal by KB which started the section in the talk page about should it be removed or not [16]. One sentence stating "A possible solution to the problem has been proposed in the GFH." and a reference citing the dissertation should not be a problem. The source is valid, the sentence is within weighting parameters, the theory exists and has been vetted by a panel of scientists in the top 10 of the field and as such I cannot see any reason for not including GFH as a possible solution to the problem.

05:45, 23 September 2010 (UTC)

Well, for one I'd like to have a link to the page that material came from, but regardless, if anything, the comments recently posted above support this consensus even more. Just to sum up the score, and please correct me if I'm wrong, because I have been wrong before, Keki Burjorjee, a Ph.D. student at Brandeis University, recently completed his dissertation on Generative Fixation, a publication he successfully defended and had approved by his advising committee, the members of whom can be found on the online abstract of his dissertation. Then he visited Wikipedia, and attempted to update the Genetic algorithm article, for whatever motive he may have had. However, one editor, a single editor, calls him out on inserting his own material into a Wikipedia article, claiming it breaks the neutral point of view Wikipedia strives to attain. And when that argument was not enough, the argument became that Ph.D. dissertations are not reliable enough for Wikipedia articles. From there Keki attempted to cull together editors in support of his views, repeatedly inserting his dissertation into the article almost as a test of consensus. And the worst part is that for the longest time I myself just sat around, listened, and agreed where necessary, such as above in my most recent reply.

I still agree with Keki that the information should be included. The argument that Ph.D. dissertations are flimsy at worst is itself a flimsy argument against the inclusion of the material in question. Making judgement on other peoples' publications based on the worst that is available makes no sense. It is not possible to judge one person's work based on another. What if this article were to be promoted to Featured Article simply because all the articles like it are pretty darn good. If anybody wants to take into question the seriousness of Keki's dissertation, I suggest they attain a doctorate in the field, read the entire paper, and then come back and present their findings. Anyway, putting that aside, it does say in WP:RS that dissertations are considered reliable sources, and for once I completely agree with the word of Wikipedia policy. While not published in a "reputable" journal, whatever that even means, a student serious about his or her education committed large amounts of time to research in a field, and put together the best summary of his or her findings as possible. Then these findings were presented to a board of experts in the field, who then looked at the paper and sent it off with a rubber stamp of approval. To say this source is not reliable is to take into question the honesty and respect of not only Keki himself, but every single person who has advised him and everybody on the committee that heard his defense. This is not a paper that one person simply wrote on his or her own time and published it to a blog or something (despite the fact that blogs have become reliable sources for a number of articles anyway), it has gone through a number of reputable people, and while it is probably true that some dissertations might not be great and have "cruft" as Looie496 said, I think that is no basis to ban inclusion of this material.

And I think the most ridiculous argument by far was that of neutral point of view. To clarify once more, Wikipedia's conflict of interest policies are merely a protection of neutral point of view. I think it is safe to say that if an author wants to insert his or her material into a Wikipedia article, the material insertion is clearly justified by Wikipedia policy, and the author has no ulterior motives for the insertion of this material, at least to the extent of our knowledge, then this material has every right to be included in the article. I think if I were to have stumbled across Keki's dissertation one day, and then come here and inserted it myself, the argument against inclusion would be half as long, probably involving only a few editors and maybe a call to RSN just to check up. What I really want to know from Glrx is exactly what reasons you have that you do not want to include Keki's material, and if it has anything to do with it being unreliable or POV, please show me how you can disprove everything everybody has said in this discussion. Furthermore, is there anybody else who opposes this inclusion that has kept quiet until now, because if so please voice your opinion now. I'd really like to know how the argument has gotten this far without consensus being established. — Parent5446 ☯ ^{(msg email)} 03:31, 14 November 2010 (UTC)

I have not followed this discussion, but anyone supporting the addition of material to an article needs to explain why that material is due and helpful (and the reference has to satisfy WP:IRS). The author of a paper is welcome to propose that information from the paper be included, and is welcome to explain how that information is helpful, but independent editors are needed to assess due/helpful/reliable. One way to proceed is to ask whether alternative reliable sources are available to support the information. If not, the information should probably not be added because Wikipedia is not the place to establish new views: it is our role to report what established science says on this topic, not to include each new view proposed in dissertations. Johnuniq (talk) 06:21, 14 November 2010 (UTC)

I suggest you check out WP:IRS#Scholarship, the guideline you just cited. Ph.D. dissertations are considered RS, and for good reason as I described above. As for whether the dissertation deserves its due weight, personally I believe so, but I do not claim to be an expert in genetic algorithms so I could be wrong. But from my perspective, even though the new theory proposed for insertion here is indeed relatively new and is without many other sources, it is still grounded in verifiable fact. We are not talking about conspiracy theories or pseudoscience, but actual research in the scientific community. While Wikipedia does try and protect from giving topics more notability than there should be, it does not attempt to filter out new scientific theories just because there are not more than X number of sources related to the theory. Furthermore, up to this point, it has only been proposed that a single sentence describing the topic be inserted. We're not making an entire section from this one new theory, which would blatantly be giving undue weight. And you are right that "Wikipedia is not the place to establish new views", and that "it is our role to report what established science says", but since when did Ph.D. research and theses outlawed from the realm of established science? — Parent5446 ☯ ^{(msg email)} 18:23, 14 November 2010 (UTC)

I've been away from WP for a while now (hence my silence). I can't claim to have read every word of the discussion that has ensued during my absence, but from my skimming, I can't see any compelling argument that would cause me to change my mind. The lack of citations, adherents, etc. mean that to mention this theory in this article would be to give it undue weight, by a mile. Yes, PhDs are peer-reviewed, but there are hundreds of thousands of such PhDs every year; peer-review on its own does not indicate that it has adherents, much less that the ideas have been accepted by the community. The bare minimum that would tickle my interest in reconsidering my viewpoint would be the existence of peer-reviewed publications by independent authors that cite this work.

Furthermore, Keki's continued lobbying for the inclusion of his work, over several months, worries me greatly. Regardless of the outcome of this debate, he should not be making these edits to the article himself. Oli Filth^{(talk|contribs)} 03:12, 21 November 2010 (UTC)

Welcome back Oli. Perhaps we can have a more responsive back and forth now. To reiterate my argument, you and Glrx are conflating the concept of weight, which applies to article content, and the concept of notability, which applies to article existence. (If you don't feel you're conflating these two concepts, please explain how you think they differ?) I readily grant that the GFH isn't currently notable (i.e. widely cited). What I don't grant is that the GFH carries no weight. Yes, scores of Ph.D. dissertations are published on the subject of genetic algorithms each year. However, none have offered a unified hypothesis for the workings of all genetic algorithms (including those using uniform crossover) that has been vetted by a committee of scholars. The GFH carries weight because in a field starved for new ideas about how all genetic algorithms might work, this hypothesis stands alone. I should add that that my thesis committee included Lee Altenberg, and Jordan Pollack, both of whom are highly regarded within the evolutionary computation community. Their signing off on my dissertation is evidence that they considered the hypothesis developed and supported therein important enough for publication. I do not claim any of my committee members as "adherents" of the GFH. But then again I do not, strictly speaking, count myself as an "adherent" of the GFH. The GFH is simply a scientific hypothesis that has been deemed publication-worthy by scientists that know a lot more about evolutionary computation, and the way science progresses than you or Glrx do. This hypothesis derives its weight from the non-existence of other published hypotheses that attempt to do what it does---provide a unified explanation for the workings of genetic algorithms, including those that use uniform crossover.

Regarding your concern about the length of this discussion, I've told you before that this in-and-of-itself isn't evidence of an ulterior motive (if I did have an ulterior motive, I probably wouldn't have made the edits under my real name). It's the principle of the matter that concerns me (Wikipedia is not citationpedia); I can be stubborn when principles I care about are violated. If you have hard evidence that I'm incapable of keeping the best interests of Wikipedia and its readers in mind, please present it. Otherwise, stop casting aspersions on my motives. Contrary to what Glrx and you keep accusing me of, I'm not trying to "push" the truth of the GFH in the article. I seek only to inform readers of the existence of a new hypothesis about the workings of all genetic algorithms that has been deemed publication-worthy by a committee of well-regarded scholars. I don't deny that there is a COI; but a COI in-and-of-itself is not sufficient grounds for exclusion. You must to provide evidence of non-neutrality. I've explained why I believe that the GFH carries weight. What, if anything, about this explanation do you dispute? Keki Burjorjee (talk) 21:51, 21 November 2010 (UTC)

Keki, I too found this posted on the mediation cabal and have read a fair bit of this long discussion. My 2 cents: if your work is as important as you say it is, then it won't be long until a secondary source says it is and then we can mention it in the article. Wnt is right that a PhD can be considered a reliable source, but I'm not happy with wikipedia being the first place for a PhD to be cited, especially if the author of it is the one who wishes to include it. Until someone does cite your work, it is best for the encyclopedia to not include it, since wikipedia is not a place to promote new thoughts. SmartSE (talk) 22:26, 21 November 2010 (UTC)

Keki, you're admitting that the only thing that distinguishes your PhD from all the others is your opinion that it's more important/groundbreaking. That's simply not a compelling argument. I'm sure the academics you mentioned have reviewed hundreds of other PhDs between them; should they all be mentioned in the article as well? Oli Filth^{(talk|contribs)} 00:11, 22 November 2010 (UTC)

SmartSE, to follow your advice would be to cede the principle I'm trying to defend, which is that weight does not equal notability. Equating weight with notability may allow domain-ignorant editors to make more snap-decisions about whether an edit stays or goes; but, I believe it would damage Wikipedia, and turn it into an extension of what's currently fashionable in the Academy. I hold that Mendel's 1866 paper on pea plants (and the now widely-acknowledged foundation of modern genetics) had weight from day one. It did not suddenly acquire weight when it was rediscovered thirty four years after it was published; it merely acquired notability. If a forum like this one had existed in his day, Mendel could probably have defended the claim that his paper had weight to the evolutionary biologists of his day; though, probably not to lay people unwilling (or unable) to get into the details.

Oli, your question " ... should they all be mentioned in the article as well?" is simple-minded and misses the crux of the matter (by a mile). I've never argued that WP:V in and of itself is reason enough for inclusion. Weight is, of course, an important consideration. For reasons I mentioned in my previous edit, I (and others on this thread) hold that my dissertation has weight. You (and others) hold that it might not have weight because it hasn't yet been cited by prominent scientists (it has, by the way, been cited by "unprominent" ones [17]). Any historian of science can provide you with numerous examples of cases where the notability of a scientific work lagged its weight by years. So your argument, to put it in your own words, is simply not compelling. Why not acknowledge that you are not qualified to adjudicate on the weight of my dissertation, and leave the matter to editors more qualified than you? Can you, at any level, appreciate how your adjudicate-at-all-costs mentality may be detrimental to Wikipedia's readers? Keki Burjorjee (talk) 07:03, 22 November 2010 (UTC)

I wish you'd stop focusing your attention on other editors' knowledge (or lack thereof) of the subject matter; it's totally irrelevant. What we are discussing here is an editorial matter. With respect, anyone above who claims that your dissertation has weight has completely misinterpreted the term; weight is determined by how wide the acceptance of the idea is; it's not a judgment call on the idea itself. Oli Filth^{(talk|contribs)} 10:30, 22 November 2010 (UTC)

I saw this at WP:MEDCAB and thought I'd take a look. I know absolutely nothing of genetics, algorithms, or combinations thereof; however, I am well-versed in Wikipedia's policies and guidelines. I gave up reading the above discussion after the eleventy-billionth word because it had already become apparent what the problem was. Essentially, I agree with SmartSE's comment above. Without passing judgement on the work of Keki Burjorjee, I think it is clear that it is desirable to see the work cited in a secondary source before it is suitable for inclusion in Wikipedia with any sort of authority. This edit does not appear to be in the spirit of the policy on citing oneself, for example. I do not see this as either a problem of weight or notability, but rather it is a problem of reliability. Until reliable, (preferably scholarly) secondary sources refer to Keki's dissertation it seems problematic to include it. Also, I think referring to other the contributions (whether comments or otherwise) as "simple-minded" is wholly inappropriate, bordering on a personal attack. None of that, please. -- Scjessey (talk) 20:24, 22 November 2010 (UTC)

At this point after all the outside comments that have come in I am more on the edge about this, so my opinion is neutral for the time being. (Just wanted to put that out there.) Also I admit that at this point I am beginning to question Keki's motives more after his most recent comments. — Parent5446 ☯ ^{(msg email)} 02:04, 23 November 2010 (UTC)

The exclusionism exhibited here is disappointing, and is, I believe, to the detriment of current readers of the article. But that's besides the point, I suppose. The rough consensus as of now seems to be that the sentence should not be included. I'll respect that. Thank you all for your involvement. Keki Burjorjee (talk) 05:55, 23 November 2010 (UTC)

Well, I believe this discussion has reached a close with consensus not to include the sentence until further secondary sources are found that support the new theory. Any objections? — Parent5446 ☯ ^{(msg email)} 22:20, 23 November 2010 (UTC)

consensus take 3

I would simply say that there is no consensus. Remember that this is not a vote!

Too many false positives has been a problem in GAs for many years and it is good to see a possible solution in GFH. I cannot read the link KB included for the IEEE paper as I am not a member but as it is a secondary published source perhaps someone else can check it out?

The main problem I feel is that KB has defended inclusion over such a long priod of time that each time a conclusiion seems to have been reached more editors returned to the isssue and the seeming concensii (lol) have fallen over. This has led, IMO, to the incorrect conclusion that KB has been pushing his opinion.

I have read KBs paper and feel that it is unfortunate that consensus to include has not been met. I do not know why Grlx did not accept mediation but I do know that the basic tenet of standing by your convictions has been admirable. KB has been through many processes to try and gain a consensus and for that I also admire his tenacity, though I would also add that the comment "simple-minded" was ill-thought out and is verging on a personal attack - remember to comment on the comments and not the editor! - I can understand the frustration of this latest reversal of a seemingly accepted edit.

Though consensus has not been reached I do still feel that GFH deserves a mention as a possible solution and would hope that if the IEEE article proves to be a good ref Grlx will accept that inclusion of a 20 or so word sentence would be ok. Remember that GFH is only a possible solution to that one problem and that would be all that the statement should say. --Chaosdruid (talk) 10:58, 24 November 2010 (UTC)

OK, so while we're still discussing I may as well note that this argument has been posted on WP:COIN and WP:FTN in order to bring in more outside opinions. Furthermore, I ask that Glrx please go to the mediation page for this argument and answer what he is looking for in a mediator so maybe we can get that moving. — Parent5446 ☯ ^{(msg email)} 02:44, 25 November 2010 (UTC)

Chaosdruid, if KB had been a long-standing positive contributor to Wikipedia, then I would have some empathy with your position. However, unfortunately, his only edits seem to be those related to this particular issue; effectively a single-purpose account geared toward self-promotion (whether intentional or unintentional). I don't believe this to be a healthy situation.

I agree that if there are at least one or two publications that discuss the topic in non-trivial detail (i.e. more than just a passing mention in a literature review), then perhaps we should reconsider. But until that situation occurs, I strongly believe we should lay this discussion to rest. Oli Filth^{(talk|contribs)} 12:29, 25 November 2010 (UTC)

Also, to add on to what Oli Filth said, Keki says that the IEEE paper cites his paper in some way. Is there any way you can prove or show us this? — Parent5446 ☯ ^{(msg email)} 17:33, 25 November 2010 (UTC)

My opening sentence "... there is no consensus" was a little unclear - I probably should have written "consensus is to do nothing" by which I was hoping to agree on "consensus is to put it in if secondary is found/has already been found (IEEE paper)"

I would also point out that KB did an amount of editing on the GA article as a whole - around six months before including GFH and the furore that started Chaosdruid (talk) 18:30, 25 November 2010 (UTC)

Oli, please stop your slander. It’s one thing to champion the cause of a group of editors that equates weight with notability, and another thing to malign someone just to close out an argument. To say that the edits are geared towards self promotion is to accuse me of putting my interests ahead of those of Wikipedia's readers. If this was true, I would not have made the edits under my name. I knew that doing so would invite scrutiny. I did so anyway because I believed that through discussions on this page, editors would see beyond the appearance of a self-serving agenda. Clearly, I was wrong. But your treatment of appearances as facts is now bordering on a witch-hunt. Chaosdruid, regarding my so called personal attack on Oli, please note that I did comment on the comment, and not the editor. Keki Burjorjee (talk) 17:34, 26 November 2010 (UTC)

Keki, I had already "closed out my argument", but then Chaosdruid made the remarks above, which I felt it important to address. Objectively speaking, it is the case that your only edits to Wikipedia have been related to this issue (or a few edits that led up to the original insertion of this material). By definition, your edits (and your associated wikilawyering beyond all reason) are very much single-purpose, the effect of which has been to promote your own work (but as I said, I don't know whether this is intentional or not). And yes, I'm aware that I could equally well be charged with "wikilawyering" here, but at least my contributions extend to more than one section of one article! If you had also made other, positive, contributions to Wikipedia, then the apperance would be very different. Oli Filth^{(talk|contribs)} 21:26, 26 November 2010 (UTC)

I kind of agree with Oli, but I also think that this currently has very little to do with the argument at hand. The question is should we include this new theory or not? And from what I can tell the general consensus is not to include the material until further secondary sources are found (the IEEE paper possibly being one of them, though we cannot know for sure yet, as Chaosdruid pointed out). Is there anybody here that can get us access to that article? (Also, it should be noted that according to Google Scholar the IEEE paper does not reference Keki's dissertation, but I do not believe this is reliable enough to rule it out.) — Parent5446 ☯ ^{(msg email)} 21:29, 26 November 2010 (UTC)

Oli, we've been over this before. A long defense of a reasonable position (weight != notability) does not constitute any kind of wrongdoing; it certainly does not merit your latest sling, "wikilawyering". If you have anything beyond the length of this discussion to back up this new charge, please present it. Otherwise, please desist. Regarding my edit pattern, I chose to contribute to subjects that I'm qualified to discuss. I haven't edited other articles because I don't feel qualified. It's as simple as that. One doesn't have to be a long standing positive contributor to make a positive contribution. Parent5446, google "IEEE Keki Burjorjee" (note that the "genoclique fixing hypothesis" was the old moniker of the "generative fixation hypothesis". They are one and the same). Also, the use of "clamping", mentioned in the abstract of the paper, is straight out of my dissertation. Keki Burjorjee (talk) 23:18, 27 November 2010 (UTC)

OK. Now I suggest we all stop posting here unless we want to start a new discussion on a new way to improve the article. Johnuniq (talk) 02:53, 28 November 2010 (UTC)

IEEE article as secondary source

We are not quite finished - there is still the outstanding issue of a possible secondary source in an IEEE article (albeit under an earlier name) which we are looking into Chaosdruid (talk) 03:43, 28 November 2010 (UTC)

Chaosdruid, thanks for your comments on my talk page. I've made the IEEE paper available for a limited time under Fair use copyright law [18]. Also, you should know that I have previously edited parts of the genetic algorithm article that have nothing to do with the GFH under the pseudonym Jasper53. I didn't make the GFH edit under this name because I didn't want to hide the COI. Keki Burjorjee (talk) 20:23, 28 November 2010 (UTC)

I confirm the above. Jasper53 (talk) 20:27, 28 November 2010 (UTC)

I would recommend declaring that account as a legitimate alternate account before somebody starts accusing you of sockpuppeting. — Parent5446 ☯ ^{(msg email)} 20:34, 28 November 2010 (UTC)

Parent5446, I don't see how the issue of sockpuppetry could arise. I haven't used the Jasper53 account to influence any discussion that I've been involved in under the KB account. Keki Burjorjee (talk) 20:41, 28 November 2010 (UTC)

I agree, but that won't stop somebody from filing an investigation now that you've come outright and said you own both accounts, especially considering people already suspect you for COI. I'm just saying to be safe it would be in your best interest to denote the association between both of your accounts. — Parent5446 ☯ ^{(msg email)} 21:14, 28 November 2010 (UTC)

Thanks for the advice. Keki Burjorjee (talk) 22:16, 28 November 2010 (UTC)

I've taken a look at the paper, and although the dissertation is indisputably listed in the references section as [6], it's not at all obvious where in the article body it's actually cited! (All searches for "6", etc. have not shown anything.) For comparison, all the other cited references are mentioned as nothing more than in passing. Oli Filth^{(talk|contribs)} 22:26, 28 November 2010 (UTC)

The absence of a citation in the body of the article was clearly an oversight. On the first page of the article, the authors state "When any set of genes gets fixed in the population, the representation of the problem space can be thought to have changed. And the new representation may contain one or more sets of genes which may not have had a detectable fitness signal in the old representation. Each time a small set of genes gets fixed, the average fitness of the population will increase by an amount that may be tiny. As the fixation of small sets of genes continues, however, these amounts will begin to add up". This is an, in places word-for-word, description of the generative fixation hypothesis. Compare it with the subsection "The Basic Idea", in chapter 4 of my dissertation [19].

Additionally, The authors' "improved genetic algorithm" (IGA) is nothing but an SGA (simple genetic algorithm) with an added tweak called "clamping" that I introduced in my dissertation to combat the issue of "mutational drag", which is another concept that I introduced. The "problem" of mutational drag and its "treatment" by clamping only make sense in the context of the generative fixation hypothesis. The authors may even have used a Matlab script that I posted on the Mathworks website in their experiments [20]. On a side note, observe that in the authors' image compression experiment, the IGA significantly outperformed the SGA (table 1). Keki Burjorjee (talk) 23:36, 28 November 2010 (UTC)

I apologize for not being an expert in the topic, but I believe this is very iffy. We cannot base a connection between two sources on only a few sentences that happen to have the same construction of prose. And while I do find it queer that Keki's paper is in the References yet it is not cited inline, we cannot assume that the authors intended to cite the paper and just overlooked it accidentally. Furthermore, why is neither "generative fixation" nor "genoclique fixing" mentioned anywhere in the paper? (Keep in mind I know very little about genetic algorithms so I'm just hypothesizing here.) — Parent5446 ☯ ^{(msg email)} 00:13, 29 November 2010 (UTC)

I do not see KB's other account as a problem with sockpuppetry. IIRC, multiple WP accounts are not forbidden. Editors (like journal reviewers) may adopt anonymity. Using multiple accounts inappropriately is forbidden. Furthermore, I do not see a COI issue with Jasper53's edits.

The recent IEEE article is not a wp:secondary source. It is not reviewing or analyzing the work. Consequently, the article does not modify my position on WP:UNDUE.

Glrx (talk) 07:02, 29 November 2010 (UTC)

If you read my other comments Glrx you'll see I was only warning Keki about multiple accounts because it is possible people would accuse him of sockpuppetry, not because I personally thought he was using his other account inappropriately. Wikipedia does allow alternate accounts in a few legitimate and appropriate cases (see WP:Multiple Accounts).

And on the topic of the IEEE article: see WP:PRIMARY. Whether the article is a primary or secondary source has nothing to do with undue weight. WP:NPOV says "Neutrality requires that each article or other page in the mainspace fairly represents all significant viewpoints that have been published by reliable sources". The IEEE paper is a reliable source, regardless of it being primary or secondary, so therefore it adds weight to GFH. Whether the IEEE paper adds enough weight to warrant inclusion of GFH in this article is another story. (And let's not forget that there is still not a definite association between the IEEE paper and GFH.) The only real difference between primary and secondary sources in the context of Wikipedia is that editors must be careful not to insert original research when citing primary sources, but in this case that is not a problem. — Parent5446 ☯ ^{(msg email)} 07:26, 29 November 2010 (UTC)

Is anybody going to respond or are we just going to ignore the IEEE article altogether? — Parent5446 ☯ ^{(msg email)} 01:25, 8 December 2010 (UTC)

I believe you already expressed it best with your comment above about it being "very iffy"! Oli Filth^{(talk|contribs)} 09:36, 8 December 2010 (UTC)

Parent5446, I can't explain the authors' oversight. However, the relevance of the GFH to the IEEE paper, far from being iffy, should be readily apparent to anyone willing to do some legwork in the service of Wikipedia's readers---by googling "mutational drag", for example (quotes included), or by comparing the following two passages (the first from the IEEE paper, the second from my dissertation). But then, this discussion has, from the start, been less about serving the interests of Wikipedia's readers, and more about what some editors consider to be their right to adjudicate on gray areas with zero knowledge of a domain, based purely on appearances. And their defense of this right, against any challenge, by all means---including slander, stonewalling, and goalpost shifting.

Grlx, the authors do more than review or analyze the GFH. They state it as a fact. Only in a very twisted view of weight would the former lend more weight than the latter. Keki Burjorjee (talk) 03:34, 9 December 2010 (UTC)

When any set of genes gets fixed in the population, the representation of the 
problem space can be thought to have changed. And the new representation may 
contain one or more sets of genes which may not have had a detectable fitness 
signal in the old representation. Each time a small set of genes gets fixed, 
the average fitness of the population will increase by an amount that may be tiny. 
As the fixation of small sets of genes continues, however, these amounts will begin
to add up.

The Basic Idea
--------------
In Chapter 3 we demonstrated that that an SGA is capable of efficiently driving 
a set of co-adapted, unlinked genes to fixation even though the fitness 
signal of this set of genes may be weak relative to the background noise. In 
driving such genes to fixation the SGA raised the average fitness of the 
population by a small amount. When a set of genes gets fixed in the population, 
the representation of the problem space can be thought to have changed. Crucially, 
the new representation may contain one or more sets of co-adapted genes which may 
not have had a detectable fitness signal in the old representation. By subsequently 
driving one or more of these sets to fixation, the SGA can once again ``change" its 
representation, and in doing so can create new small sets of coadapted genes. And 
so on.
Each time a small set of co-adapted genes gets fixed, the average fitness of the 
population will increase by an amount that may be tiny. As the fixation of small 
sets of co-adapted genes continues, however, these amounts will begin to add up. 
Based on this thought experiment, we hypothesize that adaptation in genetic 
algorithms is driven by the iterated  ``generative fixation" of small sets of 
co-adapted genes.

[burjorjee-1] Burjorjee, Keki M. (2009). Generative Fixation: A Unified Explanation for the Adaptive Capacity of Simple Recombinative Genetic Algorithms (PhD thesis). Brandeis University. OCLC 500907878.

[1]