RDA: Reciprocal Distribution Alignment for Robust Semi-supervised Learning

TL;DR

This paper introduces Reciprocal Distribution Alignment (RDA), a hyperparameter-free semi-supervised learning framework that effectively handles distribution mismatches between labeled and unlabeled data, improving robustness and performance.

Contribution

RDA is a novel distribution alignment method that regularizes classifiers without relying on class distribution assumptions, enhancing semi-supervised learning under mismatched distributions.

Findings

RDA outperforms existing SSL methods in mismatched distribution scenarios.

RDA is effective in both matched and mismatched class distribution settings.

Theoretically, RDA maximizes input-output mutual information.

Abstract

In this work, we propose Reciprocal Distribution Alignment (RDA) to address semi-supervised learning (SSL), which is a hyperparameter-free framework that is independent of confidence threshold and works with both the matched (conventionally) and the mismatched class distributions. Distribution mismatch is an often overlooked but more general SSL scenario where the labeled and the unlabeled data do not fall into the identical class distribution. This may lead to the model not exploiting the labeled data reliably and drastically degrade the performance of SSL methods, which could not be rescued by the traditional distribution alignment. In RDA, we enforce a reciprocal alignment on the distributions of the predictions from two classifiers predicting pseudo-labels and complementary labels on the unlabeled data. These two distributions, carrying complementary information, could be utilized to regularize each other without any prior of class distribution. Moreover, we theoretically show that RDA maximizes the input-output mutual information. Our approach achieves promising performance in SSL under a variety of scenarios of mismatched distributions, as well as the conventional matched SSL setting. Our code is available at: https://github.com/NJUyued/RDA4RobustSSL.