Talk:Pattern recognition

From Wikipedia, the free encyclopedia
Jump to: navigation, search
WikiProject Robotics (Rated Start-class, Top-importance)
WikiProject icon Pattern recognition is within the scope of WikiProject Robotics, which aims to build a comprehensive and detailed guide to Robotics on Wikipedia. If you would like to participate, you can choose to edit this article, or visit the project page (Talk), where you can join the project and see a list of open tasks.
Start-Class article Start  This article has been rated as Start-Class on the project's quality scale.
 Top  This article has been rated as Top-importance on the project's importance scale.

Overview section[edit]

I really like the first three paragraphs of the overview section. I'd like to remove the rest though, because it is only partially related. Less is more. Or we need more high-quality paragraphs here. I'll let this comment sink in for a while, I've done enough edits today (and tried to justify each one), let's see how many will be undone. Hobbes (talk) 20:38, 7 May 2009 (UTC)

Recognition vs. Analysis[edit]

Is there a difference between pattern analysis and pattern recognition?--Adoniscik (talk) 21:20, 28 December 2007 (UTC)

Unsupervised learning[edit]

Should this include unsupervised methods, too? E.g., clustering has major applications for pattern recognition. -- 16:55, 25 Mar 2005 (MET)

I disagree --- that material is already covered in unsupervised learning. Both supervised and unsupervised learning are covered in a high-level way at machine learning. -- hike395 17:37, 25 Mar 2005 (UTC)

Link to peer reviewed paper[edit]

Hi, I recently added some new information regarding the coparison of various classification techniques with a reference to a peer reviewed article (van der Walt and Barnard). There seems to be some controversy on this subject, the link has been removed several times. I am currently doing my PhD on this topic an I know the information is very relevant.

Is the reference and the external link suitable for this site? If not, what can I do so that this information is not repeatedly removed? -cvdwalt

As a scholar, you should know that you may give references to strengthen your claims that are detailed in your article, and not add unexpected references without any prior presentation. Your self-promotion is then, in my opinion, irrelevant both in this article and in Wikipedia. -- (talk) 14:33, 24 March 2009 (UTC)

I agree with the previous comment. Furthermore, the research mentioned is very very specific and should maybe be moved to another page that could be linked from here. I also recall several other researchers that have dealt with this topic (e.g. T.K. Ho in her ICPR2004 paper, but there are others), which IMHO also speaks for a separate page. Hobbes (talk) 20:25, 7 May 2009 (UTC)

Merging/reorganizing Pattern recognition and statistical classification[edit]

Merge with statistical classification somehow? -- hike395 08:25, 10 Mar 2005 (UTC)

Done. Made this the main article. -- hike395 15:42, 10 Mar 2005 (UTC)

Why has statistical classification been redirected to pattern recognition of all places? Statistical classification far pre-dates pattern recognition or machine learning and has uses far beyond either. I think it is more properly a category within statistics and/or cladistics/taxonomy than pattern recognition. A link to it is warranted from this page, but why the redirect? -- User:Jrbouldin

Take a look at [1], which was the old statistical classification article before the merge. It was entirely about statistical classification algorithms, which is the subject of this article and has more detail here. If you think there is more material in statistical classification beyond the material here (or in supervised learning), feel free to write something up. We can also refactor it into the right spot. If you want to write a draft (and not make an "official" article), you can write something at statistical classification/temp, and we can discuss what to do with it. It's up to you. --- hike395 June 29, 2005 22:32 (UTC)
Thank you for the clarification and other advice. I am new but will give it a shot. Seems to me the topic needs to made more general.Jeeb 1 July 2005 01:39 (UTC)

Please see Talk:statistical classification for a continuation of this discussion. Thanks! hike395

Human pattern recognition?[edit]

Numerology (for example) links here and yet this article is entirely about pattern recognition as performed by machines. Surely humans recognize patterns too. Recury 21:00, 3 July 2006 (UTC)

Yes, how about some information on patern recognition in humans? 12:39, 19 August 2006 (UTC)
To point you in the direction, an example is how we hear words in music played backwards —Preceding unsigned comment added by (talk) 01:33, 14 December 2009 (UTC)

Hidden Messages sidebar[edit]

"Hidden messages" sidebar is out of main topic of "Pattern recognition". I see no need to have in such a dominant place links to themes like Reverse speech or Numerology. I suggest removing sidebar "Hidden messages". Msm 10:51, 13 March 2007 (UTC)

No one was against my suggestion in a week time, so I remove hidden messages because of reasons written above. Msm 22:47, 20 March 2007 (UTC)
I also object to it so I posted a message on its talk page since it's back again. Let's form a consensus so we don't add/delete it repeatedly.--Adoniscik (talk) 06:35, 2 January 2008 (UTC)

Rule of three[edit]

Someone links to Rule of three (writing) in the See Also section. I see no relevance so I'll remove it. Could someone give justification here before putting it back in. --Offput 15:40, 23 March 2007 (UTC)


Huge contradiction in the introduction of the article: techniques of a sub-domain of automatic learning cannot rely on a priori knowledge! Do you know what you speak of or just quote on scholars? —Preceding unsigned comment added by (talk) 10:15, 19 September 2007 (UTC)

major rewrite[edit]

I just rewrote the article almost totally (except for a couple of sections at the end), to reflect what I believe "pattern recognition" to actually be. The former article basically just described the process of classification, with a bit of other stuff thrown in. As used in books such as Christopher Bishop "Pattern Recognition and Machine Learning" (more or less the "bible" nowadays), pattern recognition clearly refers to any sort of automatic labeling of input data: Not just classifying data into one of a fixed set of classes, but clustering the data (including learning the number of clusters from the data), regression (producing a real-valued label), sequence labeling such as using an HMM or Kalman filter, etc. Bishop does not explicitly discuss syntactic parsing, but this would clearly also qualify as a type of pattern recognition. Benwing (talk) 06:40, 5 October 2010 (UTC)

Category:Pattern recognition[edit]

I think it should be created the category "Category:Pattern recognition" to put there all types of pattern recognition, like in the Spanish edition of Wikipedia ("es:Categoría:Reconocimiento de patrones"). What do you think?--Aliuken (talk) 19:51, 31 May 2011 (UTC)

Problem statement for unsupervised learning[edit]

The student in me was pleased to find the formal problem statement for the supervised case (although I must now learn it!). Can anyone add or cite a formal problem statement for the unsupervised case? I also added the link to Deep Learning which seems to fit well under unsupervised categorisation - trust I've not stepped on anyone's toes. P.r.newman (talk) 20:10, 5 March 2012 (UTC)


In machine learning, pattern recognition is the assignment of a label to a given input value.

This is the definition of classification, and maybe of regression for a suitably broad definition of "label", not pattern recognition. Pattern recognition is much broader, to the point that it is practically synonymous with machine learning:

Pattern recognition has its origins in engineering, whereas machine learning grew out of computer science. However, these activities can be viewed as two facets of the same field, and together they have undergone substantial development over the past ten years. (p. vii)
The field of pattern recognition is concerned with the automatic discovery of regularities in data through the use of computer algorithms and with the use of these regularities to take actions such as classifying the data into different categories. (p. 1; this could be a better definition)
In other pattern recognition problems, the training data consists of a set of input vectors x without any corresponding target values. The goal in such unsupervised learning problems may be to discover groups of similar examples within the data, where it is called clustering, or to determine the distribution of data within the input space, known as density estimation, or to project the data from a high-dimensional space down to two or three dimensions for the purpose of visualization. Finally, the technique of reinforcement learning [...] is concerned with the problem of finding suitable actions to take in a given situation in order to maximize a reward. Here the learning algorithm is not given examples of optimal outputs, in contrast to supervised learning, but must instead discover them by a process of trial and error. (p. 3)

Quotes from Bishop, Christopher M. (2006). Pattern Recognition and Machine Learning. Springer.  The rest of that books uses pattern recognition and machine learning interchangeably, making no distinction between the two apart from the one I quoted (unless I missed it; it's a big book, but I read quite a lot of it). We might as well merge this article with machine learning. QVVERTYVS (hm?) 16:55, 5 May 2014 (UTC)

I've also never heard parsing being called a pattern recognition problem. Nor is it a machine learning problem; of course grammar induction is, and the two are strongly related, but the parsing phase itself is more like pattern matching than pattern recognition. QVVERTYVS (hm?) 16:58, 5 May 2014 (UTC)
While I also do think "pattern recognition" is a bit broader than machine learning (in the sense of making more use of unsupervised methods - "recognizing patterns" is less restricted to previously known patterns; on the other hand, pattern recognition is often more associated with images as in CVPR), I consider this article to be FUBAR. Let's merge contents worth rescuing into appropriate articles, and then redirect it to Machine learning (or: carefully disambiguate links to machine learning, then redirect Patern recognition to Pattern recognition (disambiguation)). (talk) 17:48, 6 May 2014 (UTC)
You seem to be thinking of "machine learning" as supervised learning only, which is how many people outside the field think of it. But if you look at the JMLR or NIPS, you'll find that they also carry publications about clustering, matrix factorization, and other unsupervised tasks. QVVERTYVS (hm?) 19:34, 6 May 2014 (UTC)
I do, and I have the impression that there are typical patterns to the unsupervised approaches published there, such as minimizing an objective function; that play a much smaller role in database-oriented communities such as KDD. KDD also carries a lot of supervised publications; they aren't disjoint. But there are patterns more common in one than the other. Microsoft academic for example also distinguishes between data mining on one hand, and has ML+PR listed as a separate subdomain. We need to explain why such a distinction is often made, and also why it isn't "just" supervised vs. unsupervised; as both subdomains have both. As Bishop noted, PR seems to come from an engineering point of view (e.g. WP:WikiProject Robotics); ML from the AI background, and data mining/KDD from a database perspective. Not sure where statistics people feel most at home... Maybe this background/focus/coloring is the main difference of these approaches. (talk) 19:55, 6 May 2014 (UTC)
I saw your new intro and it makes me want to merge with the machine learning article even more strongly. Let's leave data mining out of the equation for now.
As for statistics, there's a subfield called "statistical learning" that borrows techniques from ML and tries to give them a firm, statistical basis; Hastie and Tibshirani are the prime exponents, and boosting comes from this field. But generally, as Hinton summarized the situation, statisticians are more concerned with getting signal from small and noisy datasets, while ML people are more concerned with generalizing and squeezing performance out of large, clean datasets (i.e. robustness vs. scalability). QVVERTYVS (hm?) 11:19, 7 May 2014 (UTC)

Proposed merge with Machine learning[edit]

Pattern recognition covers machine learning topics; machine learning covers pattern recognition topics; and I haven't found any source that can really tell the two apart (except that "pattern recognition" is ambiguous and can mean the functioning of part of the brain as well, but that's not what the article pattern recognition is about). I suggest we merge the two, preferably to machine learning since that seems to be the modern name of the field. QVVERTYVS (hm?) 16:55, 21 May 2014 (UTC)

See the section #Disputed, above, for a source that discusses the (non-)difference between PR and ML (a book called Pattern Recognition and Machine Learning, appropriately). QVVERTYVS (hm?) 16:57, 21 May 2014 (UTC)