NeurIPS 2020. Comprehensive analysis of authors, organizations, and countries.

Sergei Ivanov
Criteo R&D Blog
Published in
7 min readOct 15, 2020

--

Welcome!

This post analyzes what authors and organizations publish at NeurIPS 2020 this December, similar to the analysis I did for ICML 2020.

Papers are available here. The code is available here.

Disclaimer: As before, such analysis is prone to minor errors due to how people write their names and affiliations in CMT. So this analysis is good to get global insights rather than pinpoint exact numbers❕

Okay, let’s go.

General stats

There are in total 9454 submissions of which 1900 papers were accepted, making it 20% acceptance rate.

The number of submissions continues to grow exponentially with 40% YoY. With this rate in 2025, we may have 50K submissions (good luck to reviewers 😜).

Despite the growing number of submissions, the acceptance rate remains suspiciously constant.

Organizations

Similar to ICML 2020, Google, Stanford, and MIT maintain their status of most publishing organizations.

Interestingly, among the top-10 there are now three non-US organizations: the first time Tsinghua University from China (7️⃣th place), DeepMind (9️⃣th), and University of Oxford (🔟th).

In the figure above I increment a count for each organization by one if this organization participated in writing a paper. Another way to compute the number of papers C would be to account for the number of authors in the paper N and increment C by 1/N. This is what I call the normalized number of papers.

While normalized score is lower than the original (due to collaborations with other organizations), the relative ranking for top organizations remains almost the same.

Authors

--

--

Sergei Ivanov
Criteo R&D Blog

Machine Learning research scientist with a focus on Graph Machine Learning and recommendations. t.me/graphML