Have you ever heard Jordan use Nash's approach? If you're talking about basketball, then no, but if you're talking about math, then catch it.

Some facts:

— the article is fresh, published on the arXiv in June 2024

— some of the work was done with the participation of colleagues from Meta AI

— the work was carried out with the support of the European ERC Ocean grant

The approach can be applied in ML when working with "unbalanced" datasets. A little more precisely in the part of the translation of "balanced" — where the quality is important to us not on average, but for each class (teaser in the screenshot).

A separate respect:

— the code is available on github: reds-lab/Nash-Meta-Learning

Have a nice read and use in production.

🤗 plus in karma for likes and repost

#study