Conditional mutual information for disentangled representations in reinforcement learning

M Dunion, T McInroe, KS Luck… - Advances in Neural …, 2024 - proceedings.neurips.cc
Reinforcement Learning (RL) environments can produce training data with spurious
correlations between features due to the amount of training data or its limited feature …

10 Years of Fair Representations: Challenges and Opportunities

M Cerrato, M Köppel, P Wolf, S Kramer - arXiv preprint arXiv:2407.03834, 2024 - arxiv.org
Fair Representation Learning (FRL) is a broad set of techniques, mostly based on neural
networks, that seeks to learn new representations of data in which sensitive or undesired …

Information Bottleneck Analysis of Deep Neural Networks via Lossy Compression

I Butakov, A Tolmachev, S Malanchuk… - arXiv preprint arXiv …, 2023 - arxiv.org
The Information Bottleneck (IB) principle offers an information-theoretic framework for
analyzing the training process of deep neural networks (DNNs). Its essence lies in tracking …