Adversarial Mixing Policy for Relaxing Locally Linear Constraints in Mixup

TL;DR

This paper introduces the Adversarial Mixing Policy (AMP), a novel method that relaxes the linear constraints of Mixup by adding adversarial perturbations to mixing coefficients, improving regularization especially in low-resource scenarios.

Contribution

The paper proposes AMP, a new adversarial approach to enhance Mixup by injecting non-linearity, leading to better regularization and lower error rates in text classification tasks.

Findings

AMP reduces error rates significantly compared to Mixup variants.

The method is especially effective in low-resource settings.

Experiments on five benchmarks and models confirm its robustness.

Abstract

Mixup is a recent regularizer for current deep classification networks. Through training a neural network on convex combinations of pairs of examples and their labels, it imposes locally linear constraints on the model's input space. However, such strict linear constraints often lead to under-fitting which degrades the effects of regularization. Noticeably, this issue is getting more serious when the resource is extremely limited. To address these issues, we propose the Adversarial Mixing Policy (AMP), organized in a min-max-rand formulation, to relax the Locally Linear Constraints in Mixup. Specifically, AMP adds a small adversarial perturbation to the mixing coefficients rather than the examples. Thus, slight non-linearity is injected in-between the synthetic examples and synthetic labels. By training on these data, the deep networks are further regularized, and thus achieve a lower predictive error rate. Experiments on five text classification benchmarks and five backbone models have empirically shown that our methods reduce the error rate over Mixup variants in a significant margin (up to 31.3%), especially in low-resource conditions (up to 17.5%).