As part of the Deep Learning Indaba, we have gotten a dataset of the accepted papers from NIPS 2006 - 2016 and countries of corresponding authors. We wanted to understand the scale of the non-participation of countries/regions as well as overall trends. The main post on this is now available on the Deep Learning Indaba page: Missing Continents: A Study using Accepted NIPS Papers. This post is to just add some other visualisations that might provide some more context. Its a short post.Gratitude to all of those we spoke to during this exploration and also acknowledging the work of Emily Muller who we worked with on the visualisations.
The data was prepared by NIPS. There were some challenges with using it directly. NIPS did not use ISO country names. Some countries we had to check from the original papers. Authors seem to primarily be counted agains the country of institution, but we did see a few cases where the author country was not the institution country and NIPS used the author country. We still do believe that the analysis is interesting.