NeurIPS 2019 Stats

Diego Charrez
Sep 6 · 4 min read

The Thirty-third Annual Conference on Neural Information Processing Systems (NeurIPS) is going to be held at the Vancouver Convention Center, Vancouver, Canada from December 8th to 14th, 2019.

Quoted from the acceptance decision mail sent to the authors.

We received a record-breaking 6743 submissions this year, of which 1428 were accepted (including 36 orals and 164 spotlights)

Which reflects the reputation of this conference that is growing exponentially, where the number of submissions has duplicated just from 2017 to 2019.

Last year, tickets to NeurIPS sold out in under 12 minutes so this year NeurIPS registration will change from the prior first-come-first-served model to a randomized lottery.

NeurIPS expects that about 3500 tickets will go up for sale to the general public through the lottery. From September 6th and September 20th, those wishing to attend the conference will be able to sign up for the lottery, more details can be found here.

Authors with most accepted papers

This year Sergey Levine from UC Berkley is the most contributing author with 12 accepted papers at NeurIPS 2019, closely followed by Francis Bach of INRIA -Ecole Normale Superieure with 10 accepted papers.

Notable results also by Pieter Abbeel (UC Berkeley Covariant), Ruslan Salakhutdinov (Carnegie Mellon University) and Yoshua Bengio (Mila) with 9 accepted papers each.

Although in the last graph only 6 authors with 4 accepted papers are shown there are a total of 47 authors with 4 accepted papers, the following graph show the distribution over the number of accepted papers per author where 3690 authors got one paper accepted to NeurIPS.

Institutions with most accepted papers

It was only counted at most one institutions per paper for example if a paper had five authors and all of them were from Google, that paper only counts as one paper to Google. If an author had two or more institutions all of them added to those institutions count.

Only 5 institutions have more than 75 accepted papers to NeurIPS 2019, where 2 belong to industrial affiliations (Google and Microsoft) and the other 3 belong to academic institutions from the United States.

Google, Google Brain, DeepMind 170
Massachusetts Institute of Technology 86
Stanford University 85
Carnegie Mellon University 79
Microsoft Research 76

Additional facts

Number of authors per paper

The prevailing number of authors per paper is between 2 and 5 authors, as it can be seen only 40 papers had only 1 author, on the other hand, there is just one paper with 21 authors, followed by 4 papers with 13, 14, 15, 17 authors respectively.

Words from accepted papers’ titles

The word cloud from keywords of accepted papers’ titles show the hot topics including reinforcement learning, deep learning, etc.

Most used words from the accepted papers’ titles

Disclaimer

Regarding top institutions, it was hard to match automatically institutions names so I did it manually it may not be accurate but it will give you a rough idea about the top 25 institutions with papers accepted for this NeurIPS.

Please if you find some errors just let me know in the comments although I tried my best I am a human prone to make mistakes. Thanks for reading.

I have uploaded the code for this post to a GitHub repository.

Edit, Thanks to @renato.negrinho, @elliotmeyerson, for noticing wrong results in the institution ranking graph, I made corrected them.

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade