Talk:Neural network (biology)

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

graphic list of neural networks[edit]

with explanations of each, could be used to illustrate this article

Bluerasberry (talk) 17:04, 18 January 2022 (UTC)[reply]

Merge? (again)[edit]

This article seems to deal with very similar content with Artificial Neural Network. I'm new to this discussion and couldn't find any ongoing or archived decisions on the content of these pages. But, I do feel uncomfortable with having two pages deal with the same content since it should be centralized. too_much curiosity (talk) 05:04, 15 February 2023 (UTC)[reply]

I strongly agree. I'm quite surprised, looking at these talk page archives, that nothing of the sort has happened yet. It seems like opponents of past merges want to avoid having ANNs treated as the only/main kind of NNs, but that's not an argument to keep separate articles for NN and ANN that cover the same content. It looks like some have suggested making NN more of a disambiguation page, which seems like the best solution to me. But as you say, none of those discussions seem to have come to actual decisions. I would support anyone who wants to just start gutting this page and transplanting content to ANN. If no one does, I'll give it a go when I have the time and see if anyone complains. Justin Kunimune (talk) 20:23, 11 January 2024 (UTC)[reply]
I guess I'll save you the trouble. In my opinion, there's a fundamental difference between a network formed by real neurons in an actual nervous system, and a theoretical model constructed on a computer. And I would say that, if anything, the page here comes closer to being the primary topic. I can accept that there is duplicated content that does not need to be duplicated, and can be pruned from the less appropriate page, but that doesn't mean that there should be a merge. The main problem with this page in its current form is that it says that it's about "biological neural networks" and then goes on to have large amounts of content that should be at the other page. But that's a problem with how this page was written (or copy-pasted), not with the existence of a notable standalone topic. --Tryptofish (talk) 23:38, 11 January 2024 (UTC)[reply]
Agreed on the difference between ANNs and biological neural circuitry. But, that's really why I suggested merging. I haven't encountered any neuroscience research that uses the term "neural network" or "biological neural network". From what I've seen, it just doesn't seem to be a common used term or concept that is differentiated from that neural circuit page. Is there research you're aware of that makes biological neural networks" notable and differentiate them from the neural circuit page? too_much curiosity (talk) 03:17, 12 January 2024 (UTC)[reply]
I think the idea of this being related to neural circuit is one that is worth pursuing. Please give me a little time to get those sources, but I did want to say that for now. --Tryptofish (talk) 19:21, 12 January 2024 (UTC)[reply]
Thank you for your help! And no rush, this change doesn't feel like a huge priority. too_much curiosity (talk) 21:28, 12 January 2024 (UTC)[reply]
I've looked into this some more. I'll start with the question of how this relates to neural circuits. I think there's a progression of increasing complexity, that goes from neural pathways to neural circuits to biological neural networks. In other words, a circuit can include multiple pathways. Reflex circuits, for example, can include both an afferent sensory pathway and an efferent motor pathway. A biological network, in turn, can be made up of multiple circuits. For these reasons, I'm thinking that we should continue to keep these pages separate.
Now as for biological versus artificial networks, it's definitely true that there are a lot more publications about the artificial kind. But I also think that there's enough source material out there to justify keeping this page as a standalone page (albeit one that needs a lot of improvement). Here's a fairly recent review article that explicitly treats them as separate ideas: [1]. One can see that the source also treats the biological networks as informing the artificial ones, and that's a theme I keep seeing across multiple sources. Here's a recent research article that does that: [2]. Historically, the phrase "neural network" has been used to describe networks in the brain, as in this older book: [3]. Likewise in invertebrate nervous systems: [4]. There's source material on integrating prosthetic devices with biological networks: [5]. And here are some more books over a period of time, where biological systems are treated as informing artificial ones: [6], [7], and [8]. Even when sources cover both categories, they seem to consistently treat them as two distinct things. --Tryptofish (talk) 23:19, 13 January 2024 (UTC)[reply]
I'll just note when I studied the literature would use a terminology of either "ANNs" and "BNNs" or else "neural networks" and "neuronal networks", in cases where both biological and artificial neural nets were discussed (which seemed still common for models in the late 2000s/early 2010s at least?). At the time "neural network" could mean a computational NN, a biological NN, or a theoretical model of either (such as a spin glass) or even just something neat with matrices (such as a Hopfield net). I don't know if any of this parlance is still used since a lot of these models are obsolete or turned out to be dead ends, but it's something to ask where there's still interdisciplinary work. SamuelRiv (talk) 07:25, 14 January 2024 (UTC)[reply]
Thanks for looking into that, Tryptofish! I'll admit I don't know anything about BNNs. When I suggested merging, it was under the assumption that BNNs are already covered at Neural circuit.
If BNNs are really distinct from NCs, then I would suggest instead creating a new page at Biological neural network (which currently redirects to Neural circuit), and putting all of the BNN-specific information from this page there. Neuronal network also redirects to Neural circuit, so that should be redirected to BNN. In my view, given the similarity of the terminology, this page should just be a disambiguation page, possibly with a high-level conceptual overview of the difference between ANNs and BNNs. All technical content should be moved to either ANN or BNN.
Justin Kunimune (talk) 13:05, 14 January 2024 (UTC)[reply]
Thank you SamuelRiv and Justin for adding to the discussion and thank you Tryptofish for the research!
I agree with Justin that having a disambiguation page with a high level overview might make the most sense. I do wonder where we should add comparative discussions between the two concepts. Any thoughts? too_much curiosity (talk) 19:07, 14 January 2024 (UTC)[reply]
Thanks, everyone. I can see a lot of potential moving parts here. First, I want to point out that there is currently a discussion about possibly redirecting neural circuitry at Wikipedia:Redirects for discussion/Log/2024 January 12#Neural circuitry. I'm pointing that out particularly because, when I think about what might be the most general term for disambiguating these various topics, neural circuitry could be a good candidate. All of these things we are discussing are forms of neural circuitry (but not of neural circuits, which are a subset thereof). Alternatively, we already have Neural network (disambiguation). I definitely agree that neuronal network should not redirect to neural circuit, but instead to either neural network or biological neural network. My next thought is that I'm ambivalent about whether we should title the page neural network and have biological neural network redirect to that, or the other way around. BNN is more precise, but probably a less common search term. As for disambiguation, our options are (1) a true disambiguation page, (2) a summary style page that provides an overview of the different concepts (but in text form instead of in list form), or (3) just having disambiguation at each page, whether via hatnotes, see also sections, or the main text. As I say, I think there are a lot of moving parts here, and I'm not yet sure which way is best. --Tryptofish (talk) 21:39, 14 January 2024 (UTC)[reply]
If there's not enough content for coverage of NNs as a (historical? needs a current expert in the field to weigh in imo) unified concept, then I'd support a summary-style page.
(Maybe I should rant for a policy change on the VP, but I hate disambig pages for especially technical articles. If someone is looking up the WP article for "neural network", they won't have any idea which to click on between "neural circuit", ANN, etc, no matter how good a one-sentence description one writes.) SamuelRiv (talk) 04:36, 15 January 2024 (UTC)[reply]
So as a starting point, would anyone object to me taking material relevant to ANNs and moving it to the ANN page? Probably a lot of it is already duplicated there and can be deleted. Once this is reduced to just BNN-relevant content we can decide what to do with it, but it's sounding to me like putting it at BNN and making NN a summary-style page is a good option. Justin Kunimune (talk) 00:17, 18 January 2024 (UTC)[reply]
I think moving material to ANN and deleting the duplicated parts is fine. I'm not yet sure what I think is best to do after that, but we can certainly take things one step at a time. Thanks. --Tryptofish (talk) 18:47, 18 January 2024 (UTC)[reply]

I have moved all ANN content to Artificial neural network. Depending on what we decide to do with Neural network, a small amount of it might be moved back later. However, I think for now it's best to remove it, since the article as it was was very confusing to read, due to how it mixed together information about ANNs and BNNs. The question now is whether the BNN content that's now here should be moved a new page called Biological neural network, moved to an existing page like Neural circuit, or left here (in which case this article would have to be reworded to clarify that it's just about BNNs). Justin Kunimune (talk) 15:31, 20 January 2024 (UTC)[reply]

So what's the next step here? After skimming some of the literature, I think neural networks and neural circuits are actually the same thing. Some sources try to maintain a distinction where circuits are smaller than networks, but others just use them interchangeably. At least, this article seems to be about the same thing as Neural circuit. So my vote is to move this content to Neural circuit and redirect Neural network to Neural network (disambiguation). I agree that the disambiguation page should be fleshed out to be more descriptive. I have a draft of how that could look here. Justin Kunimune (talk) 17:54, 27 January 2024 (UTC)[reply]
No, circuits and networks are two different things. --Tryptofish (talk) 00:44, 28 January 2024 (UTC)[reply]
Okay, in that case my vote is to move this to Biological neural network. In my opinion Neural network should redirect to the disambiguation page since someone searching the term could conceivably be looking for either type. Justin Kunimune (talk) 19:55, 28 January 2024 (UTC)[reply]
I just happened upon Nervous system network models, which I didn't previously know existed, but it adds a further complication. --Tryptofish (talk) 00:24, 2 February 2024 (UTC)[reply]
Oh my, that's quite the article. It sounds from the lede like it provides some broad overview of the relationship between ANNs and BNNs, but the structure and voice of the text violates so much of the manual of style that I hesitate to use any of it. I say just leave it be. Maybe someday someone will clean it up and we can link it more deliberately to and from other NN-related articles, but I for one don't want to be that person.
On that note, do you (or anyone else here) have thoughts about moving this page to Biological neural network? I suspect that many if not most people searching for Neural network are looking for ANNs, not BNNs. Moreover, there seems to be support for a broad summary comparing ANNs and BNNs, and Neural network is the most logical place to put that. Thus, I think it would benefit many readers if upon arriving at Neural network they saw a summary that links to both, instead of just the BNN article. Indeed, it seems that was the original intent of Neural network before it became overwhelmed by ANN-specific content. Justin Kunimune (talk) 18:10, 3 February 2024 (UTC)[reply]
About that network models page, I thought I would post here in case someone felt there was something worth merging. But I share your opinion that it's a mess. I'm thinking about taking it to WP:AfD. About NN versus BNN, I've been going back and forth in my mind. I won't say that I'm exactly opposed to taking the path you suggest, but I'm leaning more towards keeping the current organization where BNN redirects here and this page explains things. --Tryptofish (talk) 23:50, 3 February 2024 (UTC)[reply]
What would you say are the arguments in favor of the current organization? Justin Kunimune (talk) 20:48, 4 February 2024 (UTC)[reply]
I'll start by agreeing with you in part, by noting that the strongest argument in favor of the change is that there are a lot more sources about ANNs than about BNNs. (Without veering into original research, I cannot say why that is. Is it because ANN is really the primary topic, or because there are more primary sources where somebody wrote a paper about something they thought up for a computer program, than there are where somebody did a detailed experimental biological study?) What gives me pause is that BNN is a much less common term used in sources, than is NN. For me, that's an argument that BNN should be a redirect to NN, where the distinction between BNN and ANN is explained. And that's what we have now. I also feel that the sources tell us that NN started out as a term for BNN, and that ANNs are based upon BNNs, rather than the other way around. Thus ANN is a topic that grew out of BNN, whereas BNN is not a topic that grew out of ANN. So I'm uncomfortable about reading the sources as telling us that ANN should be the primary topic. So if a reader comes here looking for ANN, they will be taken right where they want to go. If a reader comes here looking for NN, with ANN the topic they are thinking of, they will easily find it one click away. If a reader comes here looking for NN, with BNN the topic they are thinking of, they come directly to the right place (and if they search for BNN, they get redirected here). If we have NN as some sort of DAB page, then nobody coming to that page will come directly to the page that they are ultimately looking for. --Tryptofish (talk) 23:57, 5 February 2024 (UTC)[reply]
To address your specific concerns, both BNNs and ANNs are simply called NNs in their respective fields, so that doesn't strongly favor one over the other. It is true that ANNs came from BNNs, but historical age by itself doesn't make a topic primary if the newer topic has become more prominent.
To test their relative prominence I tried searching "neural network" on DuckDuckGo. In my first three pages of results, which I think are the same as for everyone, I counted 64 pages about ANNs, 2 pages about BNNs (one of which was this Wikipedia article), and 1 about the journal Neural Networks (which covers both). Some of the ANN pages mention the term "artificial neural network", but others only use "neural network". So this makes me confident that the majority of people coming to neural network are looking for ANNs.
That said, I don't think ANN should be the primary topic because, as SamuelRiv alluded to, I think there are many readers who aren't familiar enough with either concept to know which one they're looking for, and for whom a multi-sentence disambiguating comparison/summary would be useful. So I still think a DAB page is worthwhile, even if it requires one more click of some users. I have an updated draft here of what I think that could look like. Justin Kunimune (talk) 02:46, 6 February 2024 (UTC)[reply]
I appreciate the thoughtful way you are approaching this, and I'll say again that I keep going back-and-forth in my mind about this. Perhaps some additional editors will comment here, and that could well sway my (wavering) opinion. I've looked at your sandbox draft of an overview page, and I agree that it would be the way to go, if we go that way. If I try to track where you and I agree, we agree that neither topic is really the primary topic, more like two coequal topics. We agree that a lot of readers will arrive here looking for NN as the search term, even when they are actually looking for one or the other. I'm leaning toward agreeing that, of those readers, a majority will be looking for ANN, not BNN. I'm not very bothered by an extra click for some readers, but I do think we should seek to minimize the need to click through, if we can. I'm also hesitating because BNN just isn't a widely used term. One idea that I just thought of while writing this reply is that maybe we could do it the way you suggest, but not call the page "Biological neural network". Instead, it could be "Neural network (biological)" (which is currently a redirect to "Neural network"). So there would be an overview page at Neural network (along with the DAB page at Neural network (disambiguation)), and two pages that link from that: Artificial neural network and Neural network (biological), with Biological neural network a redirect to the latter. I'd like to hear more opinions, but I think I could warm to that. --Tryptofish (talk) 22:52, 6 February 2024 (UTC)[reply]
Wow lots of interesting discussion! Sorry for jumping in after a bit of a hiatus.
Regarding the redirect/DAB discussion, I agree with Justin that "neural networks" is the WP:COMMONNAME of ANNs. From my perspective neural networks seems so commonly used for ANNs, that "Artificial Neural Network" feels like an unnatural disambiguation since it is rarely qualified as "artificial" in the computer science world. So, I'm personally comfortable with a redirect to the ANN page, but adding "not to be confused with" header on the ANN page.
I'd also support a DAB page, but prefer my previous suggestion. I do think it's important to make them coequal topics, but I wonder if common usage is more important criteria in this case.
Regarding naming, perhaps pages with the titles:
- "Neural network (machine learning)"
- "Neural network (biology)"
Although unnatural, this naming choice recognizes that the terms are used in an unqualified manner in their respective fields. Let me know what you think!
On the Nervous system network models page, woof, what an eyesore. That being said, does it contain the information we'd want to include on a BNN page? If so, that might be a helpful starting point given that the current page has little content discussing BNNs. too_much curiosity (talk) 16:16, 8 February 2024 (UTC)[reply]
Thanks. I agree with a lot of that, except that, even if "neural networks" is the COMMONNAME of ANNs, it is likewise the COMMONNAME of BNNs. I think I would be OK with your proposed pagenames, because it allows us to treat them both as a "Neural network...". That would also work well in the search box. --Tryptofish (talk) 00:00, 9 February 2024 (UTC)[reply]
I added a note on Talk:Nervous_system_network_models about the discussion here. I may work on seeing if content can be consolidated. After that, should we open a discussion a potential page move to "Neural networks (biology)"? (I am open to opening the discussion now, but I think it may be worth getting our ducks in a row first.) too_much curiosity (talk) 15:24, 9 February 2024 (UTC)[reply]
Thanks for doing that. I think it's a good idea to start by seeing if anything on the network models page can be merged into this page (with preservation of edit history). After that, I'm thinking that AfD of that page is probably more trouble than it's worth, and the page can simply be blanked and made into a redirect to whatever we end up calling this page. It would probably be best to give a bit of time to see whether any other editors have opinions one way or the other about the possible new pagenames. Personally, the more I think about it, the more I like your proposal using parentheses. --Tryptofish (talk) 21:36, 9 February 2024 (UTC)[reply]
I think the parenthetical page names are an excellent idea. I'm fully in support of moving this page to Neural network (biology) and Artificial neural network to Neural network (machine learning), but I'm also happy to wait and give others a chance to comment. Justin Kunimune (talk) 22:19, 9 February 2024 (UTC)[reply]

Summing up[edit]

I believe that we have allowed plenty of time for anyone else to comment, and that it's now reasonable to start the process of making the improvements that we have been discussing. There are an awful lot of moving parts, so I'm going to use this subsection to try to get everything summed up. (I hope I don't miss anything, but I might!)

First, I want to note that we need to be careful about preserving page editing histories. WP:Copying within Wikipedia needs to be followed.

So, for the main issue here, it seems to me that we have consensus to have two pages:

The (biology) page will contain what is now at this page, and maybe some content merged from Nervous system network models (personally, I don't relish deciding what to merge from it). The (machine learning) page will start out as a new pagename for what is now Artificial neural network.

What else have I left out? --Tryptofish (talk) 22:15, 16 February 2024 (UTC)[reply]

That sums it up for me! I think it's worth noting that we don't have to do this all at once; in particular, merging Nervous system network models will require more work than the rest. We should start by moving this page and Artificial neural network to the parenthetical pagenames, and then work on extracting content from Nervous system network models after that. Do you think it's fair to just do that first part, since we've already discussed it, or would you rather open a formal move discussion? Justin Kunimune (talk) 23:32, 16 February 2024 (UTC)[reply]
I agree with all of that. I see no need for a formal move discussion, given the discussion we've already had here. --Tryptofish (talk) 23:33, 16 February 2024 (UTC)[reply]
But as soon as those moves are made, the redirects are going to have to be retargetted, and the DAB page updated. --Tryptofish (talk) 23:36, 16 February 2024 (UTC)[reply]
Alright, I made the move and updated the redirects and DAB page. I don't have time to do more than that for now but I'll help out with the other tasks later this week. Justin Kunimune (talk) 22:47, 17 February 2024 (UTC)[reply]
Great work, thanks so much! --Tryptofish (talk) 23:44, 17 February 2024 (UTC)[reply]
Thank you! too_much curiosity (talk) 02:10, 18 February 2024 (UTC)[reply]

Top image[edit]

The output layer should have more than one node, or else it's incapable of serving the purpose of most NNs. 106.184.139.134 (talk) 01:51, 25 April 2023 (UTC)[reply]

Wiki Education assignment: Technical Writing[edit]

This article was the subject of a Wiki Education Foundation-supported course assignment, between 30 August 2023 and 7 December 2023. Further details are available on the course page. Student editor(s): RoeJogan712 (article contribs).

— Assignment last updated by RoeJogan712 (talk) 22:08, 15 October 2023 (UTC)[reply]