When Noisy Labels Meet Long Tail Dilemmas: A Representation Calibration Method

TL;DR

This paper introduces RCAL, a representation calibration method that improves learning from noisy, long-tailed datasets by modeling class representations as Gaussian distributions and recovering clean representations for better generalization.

Contribution

The paper proposes a novel representation calibration approach using Gaussian assumptions and contrastive learning, addressing noisy labels and class imbalance simultaneously.

Findings

RCAL outperforms existing methods on multiple benchmarks.

Theoretical analysis supports the effectiveness of representation calibration.

Reconstructed representations enhance model robustness and accuracy.

Abstract

Real-world large-scale datasets are both noisily labeled and class-imbalanced. The issues seriously hurt the generalization of trained models. It is hence significant to address the simultaneous incorrect labeling and class-imbalance, i.e., the problem of learning with noisy labels on long-tailed data. Previous works develop several methods for the problem. However, they always rely on strong assumptions that are invalid or hard to be checked in practice. In this paper, to handle the problem and address the limitations of prior works, we propose a representation calibration method RCAL. Specifically, RCAL works with the representations extracted by unsupervised contrastive learning. We assume that without incorrect labeling and class imbalance, the representations of instances in each class conform to a multivariate Gaussian distribution, which is much milder and easier to be checked. Based on the assumption, we recover underlying representation distributions from polluted ones resulting from mislabeled and class-imbalanced data. Additional data points are then sampled from the recovered distributions to help generalization. Moreover, during classifier training, representation learning takes advantage of representation robustness brought by contrastive learning, which further improves the classifier performance. We derive theoretical results to discuss the effectiveness of our representation calibration. Experiments on multiple benchmarks justify our claims and confirm the superiority of the proposed method.