Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
On Efficient Training of Large-Scale Deep Learning Models: A Literature Review (arxiv.org)
38 points by PaulHoule on April 10, 2023 | hide | past | favorite | 3 comments


See also "A Bibliometric Review of Large Language Models Research from 2017 to 2023"

https://arxiv.org/abs/2304.02020

The tooling on arxiv is great, but semanticscholar has some nice paper navigation features as well.

https://www.semanticscholar.org/paper/A-Bibliometric-Review-...


I hate the 2-d UMAP maps that always show those hyper-dimensional cusps though and think well-tuned tSNE is much more civilized. I was talking to the people at arXiv the other day and found out they are working on an HTML viewer for most papers that will make arXiv a much better target for linking.


I am not smart enough to get upset over UMAP vs tSNE. 3d bar charts, no error bars and unlabeled axis are still my windmills.

If you are talking to arXiv folks, please get them to fix author search, which is horribly broken for Asian names.

Take this paper for example, https://arxiv.org/abs/2304.03717 if I then click on author, "Yuanzhi Li" it does a search using "Li, Y" which reports 4768 results.

The search for the full name gives a much better 86 results.

https://arxiv.org/search/cs?query=Yuanzhi+Li&searchtype=auth...

Ideally, they would use the author's full quoted name in the search.

https://arxiv.org/search/cs?query=%22Yuanzhi+Li%22&searchtyp...

I have reported this multiple times, nothing.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: