Heavy-Tailed Universality Predicts Trends in Test Accuracies for Very Large Pre-Trained Deep Neural Networks — mentioned by spekcular

Mon, 05 Oct 2020 02:28:58 GMT

Didier Sornette, Critical Phenomena in Natural Sciences: Chaos, Fractals, Selforganization and Disorder. Springer, 2nd edition. This covers the mathematics of heavy-tailed distributions and a lot of fun applications, and also touches on some other interesting, Taleb-esque topics (fractals, renormalization group). For financial stuff in this vein (heavy tails), Bouchaud has some books with various coauthors (e.g. Theory of Financial Risk and Derivative Pricing: From Statistical Physics to Risk Management). Admittedly I have only browsed them, but what little I've read has been good. Also, on a…

Paper and discussion history. A mention is not an endorsement.

Heavy-Tailed Universality Predicts Trends in Test Accuracies for Very Large Pre-Trained Deep Neural Networks · HackerPapers discussions

Heavy-Tailed Universality Predicts Trends in Test Accuracies for Very Large Pre-Trained Deep Neural Networks — mentioned by spekcular