Freenet skips neural nets for routing: isotonic regression and empirical Bayes win

Sanity · x · 2026-11-20

Sanity notes the right ML solution often isn't a neural network. For routing in Freenet, the team uses isotonic regression, empirical Bayes, and a few other classical statistical tricks. The interesting part, he argues, is how much you can get by combining them with what you already know about the problem — no deep learning required.

Related event: Freenet Drops Neural Nets for Routing in Favor of Classic Statistics(3 posts)→

Original post →

More from Research

Research channel →