Decision Boundary-aware Generation for Long-tailed Learning

TL;DR

The paper introduces Decision Boundary-aware Generation (DBG), a framework that generates near-boundary samples to improve class separation and accuracy in long-tailed learning scenarios.

Contribution

It identifies boundary ambiguity issues in long-tailed data and proposes DBG to generate informative near-boundary samples, enhancing class separation and overall performance.

Findings

DBG improves tail class accuracy across benchmarks.

It results in more separable decision boundaries.

DBG achieves consistent performance gains with less inter-class overlap.

Abstract

Long-tailed data bias decision boundaries toward head classes and degrade tail class accuracy. Diffusion-based generative augmentation address this problem by generating additional data, while head-to-tail transfer further mitigate the generator bias inherit from long-tailed dataset. However, we show that while head-to-tail transfer helps balance the decision space of the classifier, it also induces latent non-local feature mixing that entangles inter-class features, causing decision boundary overlap and tail class distribution shift. To address this, we first identify the problem of boundary ambiguity and then propose Decision Boundary-aware Generation (DBG) framework, which promotes near-boundary representation learning by generating informative near-boundary samples. Overall, DBG rebalances the long-tailed dataset while yielding more separable decision space for long-tailed learning. Across standard long-tailed benchmarks, DBG consistently improves tail class and overall accuracy with less inter-class overlap. The code of DBG is available at https://github.com/keepdigitalabc-svg/DBG.