Institute of Fundamental Technological Research
Polish Academy of Sciences

Partners

N. Beerenwinkel


Conference papers
1.  Czyż P., Grabowski F., Vogt J., Beerenwinkel N., Marx A., Beyond Normal: On the Evaluation of Mutual Information Estimators, NeurIPS 2023, Advances in Neural Information Processing Systems, 2023-12-10/12-16, New Orleans (US), pp.1-34, 2023

Abstract:
Mutual information is a general statistical dependency measure which has found applications in representation learning, causality, domain generalization and computational biology. However, mutual information estimators are typically evaluated on simple families of probability distributions, namely multivariate normal distribution and selected distributions with one-dimensional random variables. In this paper, we show how to construct a diverse family of distributions with known ground-truth mutual information and propose a language-independent benchmarking platform for mutual information estimators. We discuss the general applicability and limitations of classical and neural estimators in settings involving high dimensions, sparse interactions, long-tailed distributions, and high mutual information. Finally, we provide guidelines for practitioners on how to select appropriate estimator adapted to the difficulty of problem considered and issues one needs to consider when applying an estimator to a new data set.

Affiliations:
Czyż P. - University of Oxford (GB)
Grabowski F. - IPPT PAN
Vogt J. - other affiliation
Beerenwinkel N. - other affiliation
Marx A. - other affiliation

Category A Plus

IPPT PAN

logo ippt            Pawińskiego 5B, 02-106 Warsaw
  +48 22 826 12 81 (central)
  +48 22 826 98 15
 

Find Us

mapka
© Institute of Fundamental Technological Research Polish Academy of Sciences 2024