Dealing with the Evil Twins: Improving Random Augmentation by Addressing Catastrophic Forgetting of Diverse Augmentations

2025-06-09

AI Summary

Random data augmentation, though computationally inexpensive, suffers from inherent stochasticity that induces “augmentation collisions,” distorting feature learning—akin to catastrophic forgetting—and severely impairing out-of-distribution generalization. This work is the first to formally interpret such degradation through the lens of catastrophic forgetting in data augmentation. We propose a lightweight forgetting-mitigation framework: (i) a feature consistency constraint to suppress interference from conflicting augmentations, and (ii) an augmentation trajectory memory mechanism to stabilize representation evolution. Our method introduces no additional parameters, requires no pretraining or architectural modifications, and seamlessly integrates with standard CNNs and ResNets. Evaluated on multiple single-source domain generalization (sDG) benchmarks, it achieves state-of-the-art performance, improving average accuracy by 3.2–5.7% over prior methods—outperforming even computationally intensive advanced augmentation techniques.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Data augmentation is a promising tool for enhancing out-of-distribution generalization, where the key is to produce diverse, challenging variations of the source domain via costly targeted augmentations that maximize its generalization effect. Conversely, random augmentation is inexpensive but is deemed suboptimal due to its limited effect. In this paper, we revisit random augmentation and explore methods to address its shortcomings. We show that the stochastic nature of random augmentation can produce a set of colliding augmentations that distorts the learned features, similar to catastrophic forgetting. We propose a simple solution that improves the generalization effect of random augmentation by addressing forgetting, which displays strong generalization performance across various single source domain generalization (sDG) benchmarks.

Problem

• Addressing catastrophic forgetting in random augmentation
• Improving generalization effect of random augmentation
• Enhancing performance in single source domain generalization

Innovation

• Addresses catastrophic forgetting in random augmentation
• Improves generalization via diverse augmentations
• Simple solution for single source domain generalization