FMMI: Flow Matching Mutual Information Estimation

2511.08552v1 cs.LG, cs.IT 2025-11-15
Авторы:

Ivan Butakov, Alexander Semenenko, Alexey Frolov, Ivan Oseledets

Abstract

We introduce a novel Mutual Information (MI) estimator that fundamentally reframes the discriminative approach. Instead of training a classifier to discriminate between joint and marginal distributions, we learn a normalizing flow that transforms one into the other. This technique produces a computationally efficient and precise MI estimate that scales well to high dimensions and across a wide range of ground-truth MI values.

Ссылки и действия