[02139] Normalizing Flows Based Mutual Information Estimation
Session Time & Room : 4E (Aug.24, 17:40-19:20) @A502
Type : Contributed Talk
Abstract : Mutual Information is a measure of mutual dependence on random quantities without specific modelling assumptions. However, estimating mutual information numerically from high-dimensional data remains a difficult problem. We propose a principled mutual information estimator based on a generalization of normalizing flows. The proposed method uses an autoregressive structure in estimating mutual information with estimating marginal and joint entropy simultaneously. Empirical results demonstrate that our proposed estimator exhibits improved bias-variance trade-offs on standard benchmark tasks.