Talk:Connectionism/Archive 1

Archive 1

Untitled

Sorry but i corrected a lot of spelling mistakes, and also substituted 'cognitivism' for 'symbolic computationalism' a phrase of which i must confess i had not heard. BScotland

PDP methods

The hyperlink associated with "Parallel Distributed Processing" under "Behaviorism" merely links back to this same page.NStone

Agree with NStone, 'connectionism' is a methodology and should separate into two categories - 'Neural Networking' and 'Parallel Distributed Processing'. NN is a method of more accurately modelling neurophysiology, whereas PDP methods are agnostic as to whether the models should reflect neurophysiology accurately or not. Therefore, 'PDP' should not redirect back to 'connectionism' in this manner.

Too Strict

Memory is created by modifying the strength or the architecture of the connections between neural units (hence the name connectionism). The connection strengths, or "weights", are generally represented as a (N×N)-dimensional matrix. Connectionists generally stress the importance of learning in their models. As a result, many sophisticated learning procedures for neural networks have been developed by connectionists. Learning always involves modifying the connection weights. These generally involve mathematical formula to determine the change in weights when given sets of data consisting of activation vectors for some subset of the neural units.

This page is too restrictive I think, connectionism does not apply solely to neural networks and training through backprop. This is one type of connectionism. For example, "learning always involve modifying connection weights" is not completely false per se, but is too restrictive of a statement, ie many algorithms change the network topology among other things. I'm not positive what exactly to change, but connectionism is a more abstract idea and should be worded as such rather than in terms of specific NN terms and methods. --Freshraisin 15:09, 1 Mar 2005 (UTC)

Fair bit of work

I have tried to address the concern that the article was too restrictive by adding quite a bit to expand on connectionism as a general framework. I have also changed "cognitivism" to "symbolism" as cognitivism is too broad and "symbolism" is often the phrase used within the debate.

Activation 'values'

From the article: "Any given mental state can be described as a (N)-dimensional vector of numeric activation values over neural units in a network." the idea of 'numeric activation values' is not mentioned before this point in the article, only the concept of 'activation'. As a result it's not clear to me what the relationship between activation and activation 'values' is. Someone who knows something please clarify this.

Suspected Excessive Promotion of Herve Abdi et al

Four obscure refs to Herve Abdi! Inserted by an anonymous user with ip address 129.110.8.39 which seems to belong to the University of Texas at Dallas. Apparently the only editing activity so far has been to insert excessive references to publications by Herve Abdi (of the University of Texas at Dallas) and his co-authors. The effect is that many Wikipedia articles on serious scientific topics currently are citing numerous rather obscure publications by Abdi et al, while ignoring much more influential original publications by others. I think this constitutes an abuse of Wikipedia. As a matter of decency, I suggest to 129.110.8.39 to remove all the inappropriate references in the numerous articles edited by 129.110.8.39 , before others do it. Truecobb 19:58, 14 April 2007 (UTC)

I have waited for a long time; now I'll delete the four obscure references. Truecobb 20:05, 15 July 2007 (UTC)

Reductionsm bad?

Reductionism is for those (us) with a positivist leaning a good thing. The claim that explaining cognition as neural firing and communication does not allow for rational thinking or feelings seem to presuppose a lot of things (antimaterialsm, soul) without any argument or reference. It is certainly not a neutral stance —Preceding unsigned comment added by 82.182.25.36 (talk) 09:12, 18 December 2007 (UTC)

-Just to address this, I think the point is roughly: If you want to know if you are missing an arm, you are better off asking an anatomist than a microbiologist. It's not that "reductionism" is bad, but rather that it is not always appropriate or useful, depending on the issue at hand. —Preceding unsigned comment added by 60.242.91.158 (talk) 18:22, 24 May 2009 (UTC) And the reference to Geake (Educational Research 2008, 50(2)) is curious: unless I've missed something, Geake makes no reference to Reductionism or Connectionism. — Preceding unsigned comment added by 220.239.216.93 (talk) 10:38, 9 January 2012 (UTC)

I agree that the reductionism comment is more than a little questionable and should be removed or altered. The defense above, even if true, does not make reductionism a bad thing that can be the basis of criticism without further support. (Let me say that the argument is not convincing to me. If you want to know why your car doesn't run or your body isn't working right--it's a good idea to consult a reductionist.) Reductionism is not good or bad from a scientific perspective (though it is generally a very productive method of analysis). So, the statement as written is meaningless and in no way enhances the article. In fact, it simply promotes a particular ideology or philosophy which as no place in this article. How helpful would if be if someone wrote, "This approach has been praised as reductionist"? Robotczar (talk) 14:40, 23 April 2014 (UTC)

Additional forms?

The article as it is now refers to multiple forms of connectionism, but doesn't even list others. Instead the only one mentioned is neural networks, and that is discussed in detail without any mention of alternative models/forms. FliesLikeABrick (talk) 13:44, 9 May 2008 (UTC)

What about HTM?

Hello, you are talking about another ways of making neural networks, one example could be the HTM, developed in Numenta. Hope it helps anybody —Preceding unsigned comment added by 190.21.45.123 (talk) 16:03, 6 January 2009 (UTC)

Related Courses

Some courses may be included to external references. —Preceding unsigned comment added by 158.195.28.144 (talk) 13:26, 18 April 2009 (UTC)

Google chrome displays a scary message when you try to click the above link. I feel like this is a peculiar bot linking sketchy websites, but who knows. 96.11.254.206 (talk) 22:30, 1 February 2021 (UTC)

"Problems" not mentioned

There are a couple of commonly cited "problems" for connectionism which I don't see mentioned here, namely 1. single trial learning, and 2. the operations over variables problem raised by e.g. Gary Marcus. Were these solved while I wasn't looking? —Preceding unsigned comment added by 60.242.91.158 (talk) 23:38, 21 May 2009 (UTC)

Second language acquisition navbox

Connectionism is quite an important concept in Second language acquisition, and I've included it in the {{SLA topics}} navbox, but it looks like putting the navbox on this page would be a bit conspicuous. Maybe this is a reason to create a new page? I'd like to hear your thoughts. — GypsyJiver (drop me a line) 04:42, 15 December 2010 (UTC)

Connectionism apart from PDP

The sentence 'Another form of connectionist model was the relational network framework...' is most unsatifying as the linke 'relational network' leads to a page that does not even contain the words 'relational network'. Not good. — Preceding unsigned comment added by 176.25.3.33 (talk) 17:10, 16 April 2012 (UTC)

Learning

I suggest changing "Connectionists generally stress the importance of learning in their models.[citation needed]" to "The weights in a neural network are adjusted according to somelearning rule or by a leaning algorithm." p.r.newman (talk) 14:35, 9 December 2012 (UTC)

Sounds good - go for it. Remember to be bold! — Mr. Stradivarius ^{(have a chat)} 15:25, 9 December 2012 (UTC)

Yes, be bold, but unless you have a source, you need to keep the [citation needed]... Lova Falk talk 21:00, 9 December 2012 (UTC)

Done (I was just passing through last time!). I replaced the needed citation with a link to the learning rule page which I think suffices. p.r.newman (talk) 15:38, 26 February 2013 (UTC)

This whole article really, really needs more inline sources. If you could find a source for your sentence or for the section, that would be great. Lova Falk talk 17:15, 26 February 2013 (UTC)