Joint Debiased Representation and Image Clustering Learning with Self-Supervision

TL;DR

This paper introduces a novel joint clustering and contrastive learning framework that employs a debiased contrastive loss to effectively handle long-tailed data distributions, improving representation learning for imbalanced datasets.

Contribution

It proposes a new debiased contrastive loss combined with divergence clustering loss to enhance clustering and representation learning on imbalanced datasets.

Findings

Improved clustering performance on long-tailed datasets

Effective handling of minority classes in representation learning

Enhanced contrastive learning with debiased loss

Abstract

Contrastive learning is among the most successful methods for visual representation learning, and its performance can be further improved by jointly performing clustering on the learned representations. However, existing methods for joint clustering and contrastive learning do not perform well on long-tailed data distributions, as majority classes overwhelm and distort the loss of minority classes, thus preventing meaningful representations to be learned. Motivated by this, we develop a novel joint clustering and contrastive learning framework by adapting the debiased contrastive loss to avoid under-clustering minority classes of imbalanced datasets. We show that our proposed modified debiased contrastive loss and divergence clustering loss improves the performance across multiple datasets and learning tasks. The source code is available at https://anonymous.4open.science/r/SSL-debiased-clustering