Separating common from salient patterns with Contrastive Representation Learning
This work addresses the challenge of learning semantically expressive representations for contrastive analysis, which is important for applications like medical diagnosis where distinguishing healthy from diseased patterns is critical, though it appears incremental as it builds on existing contrastive learning frameworks.
The paper tackled the problem of separating common from salient patterns in contrastive analysis by proposing SepCLR, a method that reformulates the task using the InfoMax principle and introduces a novel mutual information minimization strategy, achieving state-of-the-art results on six datasets including visual and medical benchmarks.
Contrastive Analysis is a sub-field of Representation Learning that aims at separating common factors of variation between two datasets, a background (i.e., healthy subjects) and a target (i.e., diseased subjects), from the salient factors of variation, only present in the target dataset. Despite their relevance, current models based on Variational Auto-Encoders have shown poor performance in learning semantically-expressive representations. On the other hand, Contrastive Representation Learning has shown tremendous performance leaps in various applications (classification, clustering, etc.). In this work, we propose to leverage the ability of Contrastive Learning to learn semantically expressive representations well adapted for Contrastive Analysis. We reformulate it under the lens of the InfoMax Principle and identify two Mutual Information terms to maximize and one to minimize. We decompose the first two terms into an Alignment and a Uniformity term, as commonly done in Contrastive Learning. Then, we motivate a novel Mutual Information minimization strategy to prevent information leakage between common and salient distributions. We validate our method, called SepCLR, on three visual datasets and three medical datasets, specifically conceived to assess the pattern separation capability in Contrastive Analysis. Code available at https://github.com/neurospin-projects/2024_rlouiset_sep_clr.