Talk:Neural coding

figure on the right?

the section on "temporal coding" refers to a figure. where is it?

watson (talk) 08:58, 18 March 2009 (UTC)

The article history shows that a bot removed two images that lacked copyright information back in December. Looie496 (talk) 16:36, 18 March 2009 (UTC)

Should temporal correlation code, in which the spike timing and the interspike timing is measured, be included? (rather than just spike timing, where the precise timing of each spike (for each neuron) is determined for each (typically 50,100,300 ms) period) seunghwane (talk) —Preceding undated comment added 23:28, 13 May 2009 (UTC).

The word 'moment' should have a link or it should be precisely defined. seunghwane (talk) —Preceding undated comment added 23:48, 13 May 2009 (UTC).

The section on the temporal code does not refer to the origins of the temporal code, but to a text book - can anybody correct this? —Preceding unsigned comment added by 130.226.230.7 (talk) 10:13, 5 October 2009 (UTC)

Merger proposal

Another editor, not me, is proposing to merge Neural coding with Neural ensemble. --Tryptofish (talk) 19:27, 6 April 2010 (UTC)

Oppose merger. I can see some opportunities to fix redundancies, but these seem to me to be sufficiently distinct topics to justify separate pages. Although I agree that most coding studies focus on ensembles, not all do, so there is some material on coding that exists outside of ensembles, and, ensembles have structural or anatomical features that are independent of coding. --Tryptofish (talk) 19:27, 6 April 2010 (UTC)

Since no one has spoken in favor of the merger, I'm going to remove the templates. --Tryptofish (talk) 19:40, 15 April 2010 (UTC)

Merger possibilities

In the "see also" section of this page, there are seven other pages, each dealing with a specific theory of neural coding. Is it a good idea to have all these separate pages, or should we consider merging all of those other pages into this one (essentially making each one a section of this page, when that section does not already exist)? I'm not yet making a formal merger proposal, but just gaging what other editors think. --Tryptofish (talk) 20:33, 22 June 2010 (UTC)

I checked and none of these seven other pages are longer than an (elaborate) section. I would support a merger proposal. Lova Falk talk 11:17, 24 June 2010 (UTC)

I agree with you about that, and I now Support such a merger. I would consider it necessary to actually incorporate material from these other pages into this page, thereby expanding this page, for the merge to be appropriate. I'm going to formally template the pages, to discuss this proposal. --Tryptofish (talk) 22:35, 16 July 2010 (UTC)

Support it's all neural coding, this is like having my grocery list on five different pieces of paper. ADDENDUM: to expand a bit, I think these are sufficiently notable topics to deserve their own page, but none of these pages are developed enough to merit that, they're basically all summaries like in the main neural coding page. But I agree with trypto that we should actually merge, not delete the others if the claims are cited or obviously true Xurtio (talk) 09:44, 10 August 2010 (UTC).

I think sparse coding can be merged into the neural codin as sparse coding is a type of neural coding and neurons are exclusively involved in the neural coding.— Preceding unsigned comment added by 129.241.134.218 (talk • contribs) 02:57, 1 December 2010 (UTC)

It's clear that the consensus supports this merge, and I apologize for being such a slowpoke, but I'll get around to it soon. Per the discussion below, I've added NeuroElectroDynamics to the proposed merger, but it should end up being no more than about one sentence here. --Tryptofish (talk) 15:58, 7 February 2011 (UTC)

Argument against a merger with Sparse coding: sparse coding and its derivatives (unsupervised learning of sparse representations) have become important in machine learning, outside of the context of neural coding and the associated biological analogies. Examples of papers in that area include (Teh, Welling, Hinton & Osindero 2003, JMLR), (Lee, Battle, Raina & Ng, NIPS'2006), (Ranzato, Poultney & LeCun, NIPS'2006), (Elard & Aharon 2006, IEEE Trans. Image Proc.), (Ranzato, Boureau & LeCun, NIPS'2007), (Grosse, Raina, Kwong & Ng, UAI 2007), (Kavukcuoglu, Ranzato & LeCun 2008, tech. rep.), (Lee, Ekanadham & Ng, NIPS 2007), (Bradley & Bagnell, NIPS'2008), David Bradley's PhD thesis (2009).— Preceding unsigned comment added by Yoshua.Bengio (talk • contribs) 00:48, 26 March 2011 (UTC)

Hmmm... In its present form, sparse coding says nothing about that. Perhaps a spinout article dealing with the machine aspects would be the way to go? --Tryptofish (talk) 18:47, 26 March 2011 (UTC)

I've got to agree with Tryptofish on this one, perhaps there can be a disambiguation page for the various "codings", since this problem will come up beyond the sparse coding issue.--Thomea neuro (talk) 16:52, 5 May 2011 (UTC)

I am also against a merger (also against a pure-ML spinoff). Indeed the Sparse Coding and Temporal Coding articles could do with being vastly extended in my opinion. I just saw that they are only about a screen or two each in length. Compare Scholarpedia for example. While Sparse Coding does play an increasing role in the ML community, its role in computational neuroscience is also still quickly increasing and is definitely underrepresented here on Wikipedia. Instead of merging, I suggest that we should expand the Sparse and Temporal Coding articles and treat them as extended subjects articles. Compare e.g. any article on "Country X" linking to the "History of Country X". — Preceding unsigned comment added by 79.204.204.143 (talk) 21:14, 7 October 2012 (UTC)

Again, my apologies for never having followed through on the merge. I think we all agree that this won't be a matter of spinning everything into machine learning. However, absent any scientific evidence that any one of these coding patterns, or any combination of them, is actually how the human brain codes anything, it would be difficult to make a case (and certainly not a case based on what another Wiki does) that some of these should be standalone pages while others should not. I agree that some topics should be given plenty of space and detail, but that doesn't mean that the space has to be with each on its own page. --Tryptofish (talk) 22:20, 7 October 2012 (UTC)

Rate coding has been merged. I will start merging the articles together one at a time. Should I not merge Sparse coding and Temporal coding? If I don't get a response, I will merge them last and anyone can feel free to undo it. --Iamozy (talk) 15:34, 3 February 2014 (UTC)

Population coding has been merged --Iamozy (talk) 15:50, 3 February 2014 (UTC)

Correlation coding and Independent-spike coding have been merged under the Population coding subsection. --Iamozy (talk) 16:07, 3 February 2014 (UTC)

Phase-of-firing code has been merged under the Temporal coding subsection. --Iamozy (talk) 16:28, 3 February 2014 (UTC)

From the "Department of Extreme WP:There is no deadline", (sorry), I've just reverted someone's removal of the remaining merge tags, because I still think that the remaining merges should eventually be performed. Arguments made in the interim notwithstanding, the fact remains that having separate pages ends up being content forks, because there is no definitive science about how coding really takes place. --Tryptofish (talk) 23:30, 1 March 2014 (UTC)

If I removed the tags in error, then I apologise. However, reading the above discussion, there was no clear consensus in over 4 years. I think it's time to either boldly do it yourself, or remove the tags. --NickPenguin_(contribs) 06:06, 2 March 2014 (UTC)

Thanks, I understand. But, as I said, there is no deadline (and, for better or, mostly, for worse, I've been preoccupied with other things). As you can see just above, another editor has been doing most of the merges, and it just happened that you came along in the middle of that process. --Tryptofish (talk) 21:06, 2 March 2014 (UTC)

Sparse coding has been merged The entire article is written in the context of the human brain and neuron coding, so I think it's relevant to add this here. I will leave the page unchanged for some time, if anyone has issues with this merge. --Iamozy (talk) 18:51, 5 March 2014 (UTC)

No issues from me. Thanks! --Tryptofish (talk) 22:17, 5 March 2014 (UTC)

Temporal coding has been merged --Iamozy (talk) 20:38, 14 April 2014 (UTC)

sparse coding should not be merged into this page. Just Google "Sparse coding", on the first page are all related to machine learning & computer vision field.

Removing new NeuroElectroDynamics section

I am for the moment removing a section on NeuroElectroDynamics that has been added. The contents I am removing are:

NeuroElectroDynamics (NED) is a model that describes brain computations in terms of universal physical principles. Memory processes, information storage are related to physical machinery that keeps information unaltered for longer time periods in the brain embedded in neurons within distributions of electric charges at macromolecular level (e.g. proteins) <ref name="Aur">Aur, D. and Jog, M.S. 2010 ''Neuroelectrodynamics- Understanding The Brain Language''. IOS Press 2010</ref>. Reading (decoding) and writing information (coding) can be simultaneously performed by electric interactions mediated by neurotransmitter release. The coding phase includes changes in spatial rearrangement of electrical charges in macromolecular formations, determined by selective gene expression, conformational changes in proteins, polarizations, while the decoding phase can be related to alterations in transient charge density dynamics. NED highlights a general physical model of computation by interaction which is a non-Turing computational model and represents an alternative to current temporal coding models.

In order for this article to include this material, there should be evidence that it has drawn attention from other workers in the field. The book is not sufficient -- there should be either reviews or publication of articles on the underlying concepts in peer-reviewed journals. Looie496 (talk) 18:19, 6 February 2011 (UTC)

The content has been re-inserted. I have restored to status quo ante. Looie496 is right. This is not a comment on the merit of the theory. Wikipedia's sourcing policy, particularly when applied to biomedical articles (see this guideline), insists that new thought may be included only when it has been evaluated by uninvolved experts in peer-reviewed journals or university-level textbooks, and when its inclusion will not give it undue weight (see this fundamental policy).

If you make a change to an article, and it is reverted, polite practice here (per this essay) is to discuss the matter on the article's talk page. Please feel free to discuss this matter here, but the discussion will revolve around those policies I've linked to, not the merits or nature of the theory. --Anthonyhcole (talk) 05:31, 7 February 2011 (UTC)

I appreciate the good faith efforts of the new editor who added the material, but I have to agree with Looie and Anthony. I also think that these concerns apply to the linked page, which is referenced entirely to the research of a single laboratory group. In the talk section above, I raised the idea of merging numerous pages, and I have to admit that I am hugely slow in getting around to acting upon it, but I think that NeuroElectroDynamics should be included in the merge as well, albeit only as a brief mention here, and mainly to make the other page a redirect. --Tryptofish (talk) 15:52, 7 February 2011 (UTC)

I feel that authors of NeuroElectroDynamics are right. How is temporal information stored in the brain? In order to replicate their measurements one has to perform invivo experiments and to insert four electrodes close to the cell. Many recordings use single electrodes. The mathematical model is one thing, their experiment shows patterns in spikes. I agree that the entire paragraph can be altered; however it should not be entirely deleted — Preceding unsigned comment added by Goodwillein (talk • contribs) 17:14, 7 February 2011 (UTC)

Additionally, there are many 'pieces of work' that can be referenced, their published papers in peer reviewed journals and the IOS book as required (see this fundamental policy). — Preceding unsigned comment added by Goodwillein (talk • contribs) 18:03, 7 February 2011 (UTC)

Concerning peer-reviewed publications, I find only a few papers in minor journals such as Neural Processing Letters and BMC Neuroscience, and even those have not been cited by anybody except Aur and Jog. I have some respect for Mandar Jog due to the very interesting work he did when he was in Ann Graybiel's lab, but I see no evidence that this particular work has garnered any attention except from the people who are doing it. Looie496 (talk) 18:51, 7 February 2011 (UTC)

Your discussion regarding authors is not relevant. Have you read their book before deleting what I inserted? Concerning peer-reviewed publications and high ISI ranking journals in biomedical field you should read http://www.nature.com/nmeth/journal/v8/n2/full/nmeth0211-104.html. I feel that would be fair to reference in neural coding all current trends in the field (Correlation coding, Independent-spike coding, Phase-of-firing code, Population coding, Rate coding, Sparse coding, Temporal coding and NeuroElectroDynamics)and see what the future holds for any of them. — Preceding unsigned comment added by Goodwillein (talk • contribs) 19:27, 7 February 2011 (UTC)

Goodwillein, I actually agree with you that all of these models should be included in this page, including (briefly) NED. The discussion here is more a matter of how much space within the page any given model should receive. Part of this is addressed in WP:UNDUE, which you linked above, and I think that the take-home message about how both Anthony and you referred to it is that there is nothing wrong (at least in my opinion) with including it, but that it is not helpful to a general public audience to devote an entire section to it, because doing so makes it seem more agreed-upon, by all of the researchers in the field, than it really is. Please also take a look at WP:N and WP:PSTS. You will see there that the way Wikipedia looks at material is somewhat different than how scientific journals do. At Wikipedia, we have to rely upon secondary sources (defined at PSTS) to indicate whether an idea from one research group has become sufficiently widely accepted that it has been taken up by other scholars, independent of those who first proposed it. --Tryptofish (talk) 20:08, 7 February 2011 (UTC)

Tryptofish, I also agree with you, please select what you would like to see from what I inserted that can be (briefly) added. — Preceding unsigned comment added by Goodwillein (talk • contribs) 20:26, 7 February 2011 (UTC)

Thanks. Per #Merger possibilities just above, I'll give that a try, but when I get around to it, so not today. --Tryptofish (talk) 20:31, 7 February 2011 (UTC)

Tryptofish, I think that a better solution would be to add here what is specific in each case for neural coding and let their pages unchanged since all seem to add different relevant details that cannot be included here. In case of NeuroElectroDynamics specific for coding and decoding information would be: "The coding phase includes changes in spatial rearrangement of electrical charges in macromolecular complexes of neurons, determined by selective gene expression, conformational changes in proteins, polarizations. The decoding phase can be related to alterations in transient charge density dynamics" If you agree I can add. Thanks — Preceding unsigned comment added by Goodwillein (talk • contribs) 15:25, 8 February 2011 (UTC)

Concerning Looie496 position regarding neuroscience labs ( including Graybiel's lab) we should see the difference between experimental data processing and computational theoretical models. Few neuroscience labs have generated new computational theories! Importantly, good statistics on a wrong model (temporal coding model) does not make the science better doesn’t matter where is published. Experimentally, we are at the beginning of understanding the effect of electric field (endogenous field, ephaptic coupling) in computation. Regarding the second subject neural coding and NED are not equivalent, temporal coding models can be derived as particular examples of NED with no real predictive power, semantics or significance.Giovannistefano35 (talk) —Preceding undated comment added 18:30, 29 May 2011 (UTC).

Giovannistefano I completely agree. I just added that link to show that bad statistics is published in high-impact journals, however indeed, statistical significance doesn't matter if the basic model of temporal coding is an epiphenomenon — Preceding unsigned comment added by Goodwillein (talk • contribs) 17:57, 30 May 2011 (UTC)

"Neuronal code"?

Is "Neuronal code" yet another term that belongs here? Any difference with "neural code"? Cf. Synapses, Neurons and Brains | Coursera ★NealMcB★ (talk) 05:25, 7 January 2013 (UTC)

No difference. The term "neuronal code" is not used very often, though. Looie496 (talk) 07:32, 7 January 2013 (UTC)

Good to know. And I created redirects from neuronal code and neuronal coding to this article. Lova Falk talk 18:08, 8 January 2013 (UTC)