Wikipedia:Articles for deletion/Ensmallen (C library)
Appearance
- The following discussion is an archived debate of the proposed deletion of the article below. Please do not modify it. Subsequent comments should be made on the appropriate discussion page (such as the article's talk page or in a deletion review). No further edits should be made to this page.
The result was delete. Down-weighting comments from users who appear to have a connection with the subject. -- RoySmith (talk) 16:34, 23 February 2019 (UTC)
[Hide this box] New to Articles for deletion (AfD)? Read these primers!
- Ensmallen (C library) (edit | talk | history | protect | delete | links | watch | logs | views) – (View log · Stats)
- (Find sources: Google (books · news · scholar · free images · WP refs) · FENS · JSTOR · TWL)
No coverage in independent reliable sources; searching online and in academic databases, the only coverage of the subject appears to be published by the library authors themselves. signed, Rosguill talk 18:55, 7 February 2019 (UTC)
- Note: This discussion has been included in the list of Mathematics-related deletion discussions. Coolabahapple (talk) 21:47, 7 February 2019 (UTC)
- Note: This discussion has been included in the list of Software-related deletion discussions. Coolabahapple (talk) 21:47, 7 February 2019 (UTC)
- Delete At best, this is too soon: the software was presented in an arXiv preprint just last October, far too recently for it to have had significant community uptake or influence. To see the kind of sourcing this would need, compare the articles on PyMC3 and PyTorch, which were nominated for deletion and survived. XOR'easter (talk) 20:46, 10 February 2019 (UTC)
- Keep There are multiple reasons for keeping this page as is: (1) the associated paper is more than just an arXiv pre-print; it is actually published at an official NeurIPS workshop (see accepted papers and associated DOI); NeurIPS is a well-established and highly influential conference in machine learning and artificial intelligence. (2) according to Google Scholar, there are already 2 other papers which cite the associated paper, meaning that it already has impact; the citation count is only going to increase. (3) the ensmallen library is used by mlpack, a well-established library of machine learning algorithms (use evidenced by source code on GitHub: FindEnsmallen.cmake, lmnn.hpp, etc). (4) the ensmallen library implements many optimization algorithms directly described on Wikipedia, such as the seminal Stochastic gradient descent and its offspring, meaning it has direct practical consequences Gtfjbl (talk) 16:00, 13 February 2019 (AEST)
- Two citations is pretty low. Once the paper gets many citations, if the paper gets many citations, an article would be in order. As for the other arguments, I don't think that notability should be inherited between the subjects described in points 3 and 4, I think it's entirely possible for major machine learning libraries to be bloated and conversely implementing stochastic gradient descent is not unique and I don't see why that would .immediately make something likely to have received coverage. I can't say I'm familiar with the NeurIPS conference, and it does get a fair amount of coverage in the press, but from working in an adjacent field my impression is that even the best conferences are still full of tons of papers many of which don't really go anywhere or influence much. signed, Rosguill talk 06:42, 13 February 2019 (UTC)
- According to Google Scholar, the NeurIPS conference is ranked #1 in the Artificial Intelligence field, with a h5-index of 134 and h5-median of 221 (source). As such it would be safe to say it is very influential venue. One of the primary reasons for the existence of the ensmallen library is to concretely address the serious limitations of previous implementations of many optimization algorithms (including SGD and Quasi-Newton) such as brittleness, inflexibility, lack of integration capability, slow execution, etc. Since optimization algorithms are widely used in machine learning, having very robust (from multiple view points) and an unencumbered open source implementations of such algorithms has direct practical consequences for the wider application of machine learning in many products and industrial processes. Gtfjbl (talk) 17:20, 13 February 2019 (AEST)
- Being accepted by a selective conference is like getting published in a peer-reviewed journal: it's a start, but by itself, it doesn't automatically make the topic notable. We need indicators of impact and influence, like discussions of and citations to that work. The only GS citations to the paper are by its authors themselves. There's nothing wrong with that, but it is a signal that it's too soon to have an encyclopedia article about it. XOR'easter (talk) 15:56, 13 February 2019 (UTC)
- According to Google Scholar, the NeurIPS conference is ranked #1 in the Artificial Intelligence field, with a h5-index of 134 and h5-median of 221 (source). As such it would be safe to say it is very influential venue. One of the primary reasons for the existence of the ensmallen library is to concretely address the serious limitations of previous implementations of many optimization algorithms (including SGD and Quasi-Newton) such as brittleness, inflexibility, lack of integration capability, slow execution, etc. Since optimization algorithms are widely used in machine learning, having very robust (from multiple view points) and an unencumbered open source implementations of such algorithms has direct practical consequences for the wider application of machine learning in many products and industrial processes. Gtfjbl (talk) 17:20, 13 February 2019 (AEST)
- Two citations is pretty low. Once the paper gets many citations, if the paper gets many citations, an article would be in order. As for the other arguments, I don't think that notability should be inherited between the subjects described in points 3 and 4, I think it's entirely possible for major machine learning libraries to be bloated and conversely implementing stochastic gradient descent is not unique and I don't see why that would .immediately make something likely to have received coverage. I can't say I'm familiar with the NeurIPS conference, and it does get a fair amount of coverage in the press, but from working in an adjacent field my impression is that even the best conferences are still full of tons of papers many of which don't really go anywhere or influence much. signed, Rosguill talk 06:42, 13 February 2019 (UTC)
- Draftify or userfy per above. I agree that this is WP:TOOSOON, but given that it was recently released, as well as being featured in NeurIPS, suggests that more references could emerge. --DannyS712 (talk) 20:49, 13 February 2019 (UTC)
- Keep. ensmallen has actually been around for quite a while, just not as its current name. If you take a look at the paper it is the optimization framework from mlpack, which has been used in over 100 scientific publications and has well over 100 contributors on Github. So despite the fact that the NeurIPS workshop paper was published in only October, based on the git history the ensmallen codebase has existed since 2010, and the community is much larger than a single recent workshop paper would imply (see also that ensmallen itself has 60 contributors already, a sign that this is not a small or very recent effort). So personally I don't think this is WP:TOOSOON and I think it would be useful and relevant to the machine learning community on Wikipedia to keep it here. --Nemarts (talk) 14:36, 14 February 2019 (UTC)
- Question Wouldn't it make more sense to cover it within the article mlpack, then? XOR'easter (talk) 16:08, 14 February 2019 (UTC)
Relisted to generate a more thorough discussion and clearer consensus.
Please add new comments below this notice. Thanks, Ad Orientem (talk) 00:31, 15 February 2019 (UTC)
Please add new comments below this notice. Thanks, Ad Orientem (talk) 00:31, 15 February 2019 (UTC)
- Delete as not independently notable; hasn't received coverage in multiple secondary sources, just their own original workshop paper and then citing that paper in a couple of Arxiv preprints with overlapping author lists. Sneftel (talk) 09:47, 15 February 2019 (UTC)
- The above discussion is preserved as an archive of the debate. Please do not modify it. Subsequent comments should be made on the appropriate discussion page (such as the article's talk page or in a deletion review). No further edits should be made to this page.