Boosting Model Resilience via Implicit Adversarial Data Augmentation

TL;DR

This paper introduces a novel data augmentation technique that enhances model resilience by using adversarial and anti-adversarial feature perturbations, improving performance across various biased learning scenarios.

Contribution

It proposes a new augmentation method based on adversarial feature perturbations and a meta-learning framework to optimize classifiers without explicit augmentation.

Findings

Achieves state-of-the-art results in biased learning scenarios

Demonstrates broad adaptability across tasks

Provides theoretical insights into augmentation effects

Abstract

Data augmentation plays a pivotal role in enhancing and diversifying training data. Nonetheless, consistently improving model performance in varied learning scenarios, especially those with inherent data biases, remains challenging. To address this, we propose to augment the deep features of samples by incorporating their adversarial and anti-adversarial perturbation distributions, enabling adaptive adjustment in the learning difficulty tailored to each sample's specific characteristics. We then theoretically reveal that our augmentation process approximates the optimization of a surrogate loss function as the number of augmented copies increases indefinitely. This insight leads us to develop a meta-learning-based framework for optimizing classifiers with this novel loss, introducing the effects of augmentation while bypassing the explicit augmentation process. We conduct extensive experiments across four common biased learning scenarios: long-tail learning, generalized long-tail learning, noisy label learning, and subpopulation shift learning. The empirical results demonstrate that our method consistently achieves state-of-the-art performance, highlighting its broad adaptability.