Decoupled Training for Long-Tailed Classification With Stochastic Representations

TL;DR

This paper introduces a decoupled training approach for long-tailed classification, utilizing stochastic representations and SWA techniques to improve generalization and classifier robustness, validated on multiple benchmark datasets.

Contribution

It proposes a novel classifier re-training algorithm using stochastic representations from SWA-Gaussian and self-distillation, enhancing long-tailed classification performance.

Findings

Improved accuracy on CIFAR10/100-LT, ImageNet-LT, and iNaturalist-2018.

Enhanced uncertainty estimation capabilities.

Demonstrated effectiveness of stochastic representations in classifier robustness.

Abstract

Decoupling representation learning and classifier learning has been shown to be effective in classification with long-tailed data. There are two main ingredients in constructing a decoupled learning scheme; 1) how to train the feature extractor for representation learning so that it provides generalizable representations and 2) how to re-train the classifier that constructs proper decision boundaries by handling class imbalances in long-tailed data. In this work, we first apply Stochastic Weight Averaging (SWA), an optimization technique for improving the generalization of deep neural networks, to obtain better generalizing feature extractors for long-tailed classification. We then propose a novel classifier re-training algorithm based on stochastic representation obtained from the SWA-Gaussian, a Gaussian perturbed SWA, and a self-distillation strategy that can harness the diverse stochastic representations based on uncertainty estimates to build more robust classifiers. Extensive experiments on CIFAR10/100-LT, ImageNet-LT, and iNaturalist-2018 benchmarks show that our proposed method improves upon previous methods both in terms of prediction accuracy and uncertainty estimation.