Talk:Computational neuroscience

From Wikipedia, the free encyclopedia
Jump to: navigation, search
          This article is of interest to the following WikiProjects:
WikiProject Neuroscience (Rated B-class, High-importance)
WikiProject icon This article is within the scope of WikiProject Neuroscience, a collaborative effort to improve the coverage of Neuroscience on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
B-Class article B  This article has been rated as B-Class on the project's quality scale.
 High  This article has been rated as High-importance on the project's importance scale.
 
WikiProject Computer science (Rated C-class, Mid-importance)
WikiProject icon This article is within the scope of WikiProject Computer science, a collaborative effort to improve the coverage of Computer science related articles on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
C-Class article C  This article has been rated as C-Class on the project's quality scale.
 Mid  This article has been rated as Mid-importance on the project's importance scale.
 
WikiProject Engineering  
WikiProject icon This article is within the scope of WikiProject Engineering, a collaborative effort to improve the coverage of engineering on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
 ???  This article has not yet received a rating on the project's quality scale.
 ???  This article has not yet received a rating on the project's importance scale.
 
WikiProject Psychology  
WikiProject icon This article is within the scope of WikiProject Psychology, a collaborative effort to improve the coverage of Psychology on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
 ???  This article has not yet received a rating on the project's quality scale.
 ???  This article has not yet received a rating on the project's importance scale.
 
WikiProject Systems (Rated B-class, High-importance)
WikiProject icon This article is within the scope of WikiProject Systems, which collaborates on articles related to systems and systems science.
B-Class article B  This article has been rated as B-Class on the project's quality scale.
 High  This article has been rated as High-importance on the project's importance scale.
Taskforce icon
This article is within the field of Cybernetics.
 
WikiProject Computational Biology (Rated C-class, Top-importance)
WikiProject icon This article is within the scope of WikiProject Computational Biology, a collaborative effort to improve the coverage of Computational Biology on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
C-Class article C  This article has been rated as C-Class on the quality scale.
 Top  This article has been rated as Top-importance on the importance scale.
 
WikiProject Spoken Wikipedia
WikiProject icon This article is within the scope of WikiProject Spoken Wikipedia, a collaborative effort to improve the coverage of articles that are spoken on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
 

refactoring[edit]

I will soon call for colleagues fora refactoring of the page as a portal, please contact me if intersted Meduz (talk) 10:26, 4 May 2008 (UTC)

Computational[edit]

I thought "computational" was meant in the sense of "looking for computations in the brain" and not "using computers to model the brain". the last option is certainly something of little future since the field is studying the interactions of multiples agents at different levels (synapses, neurons, assemblies, areas, ...). Meduz (talk) 09:01, 8 January 2008 (UTC)

I've fixed some problems with the entry and I'll soon add content to all of the subsections. Also can someone fix the "categories page"? I don't know how to do it. thx --sluox

Computational neuroscience is more of a general descriptor of a branch of science that uses computational and mathematical methods for examine the processes of the brain, whereas neuro cybernetics is more specifically about designing interfaces. I believe neurocybernetics should be combined with Brain-computer interface instead.

Semiconscious 19:22, 22 Jun 2005 (UTC)

Another possible definition is along the lines of the division between a theoretical physicist and an experimental physicist. Thus the term 'computational neuroscience' becomes 'theoretical neuroscience'. The justification for this? Simply that the computer is merely a tool which allows for the testing of neuroscientific hypotheses, the models are derived theoretically from both consideration of the data gathered by the 'experimental neuroscientists' and from the insight (mathematical or otherwise) into the nature of neural computation provided by the 'computational neuroscientists'.

In addition, the field is also populated by systems engineers and electronics engineers.

There is a reference to a work by Shlens, et al., for which there is no full citation.

I agree that the definition of computational neuroscience is too narrow. Computational neuroscience in my opinion is far more than using computers to build toy models of neurons. Sure, that is part of it, but only a part. More generally I would say that computational neuroscience is a paradigm in neuroscience that views the brain as an information processing machine (a computer if you will). Often people studying this paradigm use models (both computational AND mathematical) encompassing many levels of description (eg, low level sub cellular, cellular, network or even abstract: for example probabilistic models or temporal difference learning etc...). Also I would say that one can be a computational neuroscientist and be an experimentalist (which is probably pretty much what a 'systems' neuroscientist is). To say that computational neuroscience is just about building twitchy Hodgin Huxley models of cells is a huge misunderstanding of the field as a scientific movement.

neuroinformatics[edit]

I think this article should not be merged with neuroinformatics, as computational neuroscience deals with computational models for various levels of neuroscience and is hence more theoretical, on the other hand neuroinformatics deals with software applications and projects used to build these, and is hence more practical. Computational models are not always necessarily made using a neuroinformatics tool. 82.6.110.89 07:50, 16 May 2007 (UTC)

Please contribute to the whole brain emulation article[edit]

Especially the Current research section. It needs sources discussing mapping and simulation of animal brains such as C. elegans. Mange01 (talk) 21:39, 2 April 2009 (UTC)

Merger proposal[edit]

Theoretical neuroscience is not Computational neuroscience[edit]

... but computational neuroscience is a branch of theoretical neuroscience. The hierarchy I've always understood was that Theoretical neuroscience was just that. A theoretical approach to neuroscience. Theoretical neuroscience, however, encompasses two subjects: mathematical and computational modeling. I've seen this differentiation made in a class syllabus , but also see:

https://redwood.berkeley.edu/bruno/papers/TN-review.pdf

"There tend to be two camps in the field of computational neuroscience, and they are probably best exemplified by how they use the term ‘‘computation.’’ In one, mathematical models are constructed primarily to describe or characterize existing data, and computation is used mainly as a means to simulate or analyze the data, in rather the same way as computational chemistry or computational fluid dynamics. In the other camp, computation is applied in a more theoretical manner, as a metaphor for what the brain is actually doing."

And see my talk page for a discussion by someone who studies the same subject I do ("neurodynamics" is a popular name in Google Scholar that explicitly describes the field, though I've never heard it before. I've heard "dynamical neuroscience" but you can see the problems that caused above. Anyway, it conforms to that mathematical branch of theoretical neuroscience. I'm a grad student, so a PhD that studies the mathematical approaches would have the highest value input, imo.

I think there should be a general Theoretical neuroscience page that differentiates mathematical from computational methods. Xurtio (talk) 06:07, 14 August 2010 (UTC)

I basically agree with the conclusion but would like to adjust the explanation a bit. "Computational neuroscience" actually comprises two quite distinct things: (1) the study of how neurons compute, (2) the use of computers to study the nervous system. The first is a branch of theoretical neuroscience, the second is not necessarily. Looie496 (talk) 19:01, 14 August 2010 (UTC)
what you've said here, I agree with; but in the context of theoretical neuroscience, the (2) is necessarily theoretical. But more than that... what about theoretical neuroscience methods that don't require a computer? Are they suddenly not "theoretical neuroscience" despite being theoretical and neuroscience? Xurtio (talk) 03:35, 16 August 2010 (UTC)
Is #2 above like when a neuroscientist uses a computer to analyze an EEG or MRI, or maybe when they use one to get spectra of neuron spikes from a rat brain probe? Those are definitely not "computational neuroscience" by any definition I'd ever hear - those are clinical and/or experimental. Let's not get too beyond mainstream here. Anyway, when it comes to mathematical models and optimization of algorithms, theoretical and computational neuroscience are identical, and I don't think they are anywhere near the academic segregation of, say, how theoretical vs. computational physics/mathematics are split. SamuelRiv (talk) 20:02, 18 August 2010 (UTC)
Consider for instance, the Hodgkins-Huxley model that were nonlinear differential equation models of neurons before computers were around. Theoretical analysis of dynamical systems requires a lot of mathematical background and despite being nonlinear, there were practical methods conceived for finding out properties of the system (Strogatz, "Nonlinear Dynamics and Chaos" is the modern text on the subject). Even though no solutions were found. (2) refers to this very same analysis, only nowadays you have the luxury of using a computer instead of paper and pencil. It's a matter of convenience. Computational theory, on the other hand, is literally based on the computational properties of neurons. Steroetypically speaking, and over-generalizing, the one (2) is a physicist/mathematical biology approach, the other (1) is a computer science/electrical engineer approach. AI is generally pursued by (1), not (2). (2) is generally a reductionist method that's more useful to physiology and medical applications, (1) is more the study of information flow regardless of the physics. These are generalizations of course( i.e Karl Friston, who utilizes the concepts of thermodynamics to make a holistic model of the brain- "Free Energy Principle of the Brain"). Jack Cowan (mentioned as a forefather in "Theoretical Neuroscience" is one of the people who made the distinction clear:
One of the major differences seems to be that (1) makes interpretations from cognitive psychology data, (2) uses experimental data from the the physics of the neurons alone.
As an alternative, we could at least break up the article Computational neuroscience into two sections (within the same page): mathematical methods and computational methods. That's how the canonical textbook, "Theoretical Neuroscience" is broken up, and it appears to be the same subject matter.
Xurtio (talk) 05:10, 19 August 2010 (UTC)
Another option (or a first step) is to create a short section in the Neuroscience article, summarizing these issues, but leave this article as it is. The section might be entitled "Mathematical models" (which I prefer) or "Theoretical and computational neuroscience".
If the mathematical models section becomes large, a separate article on Mathematical models in neuroscience can be created.
Can you give an example of a mathematical model that is not used in computational neuroscience, perhaps because it is too simple or idealized? Generally, in various computer simulation and numerical computation application areas, you tend to use more accurate and advanced models than in analytical math. However, most mathematical models can be used both in numerical computing and mathematical analysis, so it is really hard do draw a firm line between them. It is normally easier to differentiate mathematical models from simulation tools and simulation results.Mange01 (talk) 08:58, 19 August 2010 (UTC)
Your last sentence is actually my point. Computational models that have nothing to do with mathematics are the "computational approach" while mathematics models, even if they use computers, are the "mathematical approach". As Looie said, the problem is what "computation" means. 1) computer as a tool, 2) computational theory. When you're a (1) doing mathematical biology, think you're a (2) when you talk about "computational neuroscience". Originally, it would have meant the same thing as "computational physics", i.e., (1), computation is a tool to study your mathematical models. Unfortunately, (2) uses the same name, and then Abbot put the (1) and (2) in a book together called, "Theoretical Neuroscience", and some scientists ARE both (1) and (2).z
As for Mathematical models in neuroscience we currently have Biological neuron models, but it just gives the specific models, it doesn't talk about the field in general. You're other suggestion hough, to add a section explaining the difference is a good idea. And possibly my above explanation (with valid citations/research) could be a section in the Computational neuroscience article. Thanks for your input. Xurtio (talk) 01:48, 20 August 2010 (UTC)
Distinguishing a field that uses the computer as a "tool" versus using it as an intrinsic mode of study is like calling a carpenter who uses a power drill a "machinist". Everybody uses computers to investigate every mathematical model and every iota of quantitative data. The difference comes when the model and the computer simulation are intrinsically linked. For example, some mathematical models only produce useful results when integrated and relaxed over an enormous number of points (ie Lattice QCD), so deriving and manipulating those equations would still be a computational methodology. There is just no such thing, in current research, as a mathematical investigation without using a computer, or a computational investigation without using mathematics. But as in physics, there are lines at which the methodologies can be separated into different fields - my contention is that this separation does not yet exist in theoretical neuroscience without, as in using ANNs to analyze handwriting, ceasing to be neuroscience. SamuelRiv (talk) 06:16, 21 August 2010 (UTC)
By your analogy, I think you're missing the point. We're talking about a theory called "computational theory" that encompasses a philosophical ideology about how neurons work (see the first sentence of this wiki article) vs. a computational tool (for ANY theoretical approach). For the carpenter, it's as if we were saying pythagoras theorem and the power drill are the same thing to make the analogy proper. Carpenters use math, too. But, I mean, so does everybody. We can all count to 100. But a mathematical model is different than a computational model that uses math. And of course, to confuse the issue, they're not mutually exclusive, either.
I mean, isn't that why the distinction is made in Dayan's book? https://redwood.berkeley.edu/bruno/papers/TN-review.pdf Xurtio (talk) 07:58, 21 August 2010 (UTC)
Okay then, you are suggesting that comp. neuro. encompasses the philosophy that the tool is analogous to the real-life mechanism. This seems to be primarily suggesting that the computational neuroscience is about connectionism - that a ANN model reflects the reality of how neurons actually physically arrange, as opposed to being an analogy of how the computation can work. In this way, my previous statement that such a philosophy in exclusion to other approaches then "ceases to be neuroscience", as function/result supersedes the physical model (such as a biologically-motivated H-H circuit network).
At my undergrad,incidentally, the respective courses were called "Computational Intelligence" for ANNs and "Computational Neuroscience" for H-H etc, each with various degrees of exclusion (one CN prof worked with ANNs in robotics, while the other CN prof eschewed ANNs, and the CI prof hated biology). SamuelRiv (talk) 17:38, 22 September 2011 (UTC)

Cruft[edit]

I have a strong urge to de-cruftify this article, which has become essentially a link page. Any objections to doing that? Looie496 (talk) 15:14, 7 September 2011 (UTC)

Okay, I've just done a thorough housecleaning. There may still be a little bit of cruft left, but I think I got rid of most of it. Looie496 (talk) 16:43, 21 September 2011 (UTC)