Apr 3, 2025
Disclaimer: This project is intended as the author's copium. It has dubious scientific merit at best.
You can find a preprint paper of the methodology on arXiv. The result is also on an interactive website here. The scripts used to generate the result can be found on GitHub.
According to CSRankings:
...(Measuring publications in top-tier conferences) is intended to be difficult to game, since publishing in such conferences is generally difficult: contrast this with other approaches like citation-based metrics, which have been repeatedly shown to be easy to manipulate.
Obviously it is impossible to tell how much effort is being spent in each area. Instead, I am going to make the (bold) assumption that each faculty spends the same amount of effort in publishing papers. Therefore, the total amount of effort in one area is in proportion to the total number of faculties in that area. I will use the CSRankings faculty data for the purpose. If a faculty appears in multiple areas, I will assume that he spends equal effort in each area.
I use data from CSRankings, which comes from DBLP. I include papers published in 2019–2023.
ICLR is one of the most reputable conferences in machine learning. Its review process contains many entertaining and thought-provoking exchanges. I will list a few of my favorites below.
It seems like a very good time to invoke Goodhart's law and go all in on Bioinformatics:
When a measure becomes a target, it ceases to be a good measure.
Protein folding here I come!