Model Debiasing by Learnable Data Augmentation

TL;DR

This paper introduces a novel two-stage data augmentation method to improve neural network generalization on biased datasets, effectively reducing reliance on spurious correlations without requiring bias annotations.

Contribution

The work presents a bias-agnostic data augmentation pipeline that identifies biased samples and enhances model robustness, outperforming existing methods on synthetic and real datasets.

Findings

Achieves state-of-the-art accuracy on biased datasets

Improves generalization regardless of bias level

Robust performance on both synthetic and real-world data

Abstract

Deep Neural Networks are well known for efficiently fitting training data, yet experiencing poor generalization capabilities whenever some kind of bias dominates over the actual task labels, resulting in models learning "shortcuts". In essence, such models are often prone to learn spurious correlations between data and labels. In this work, we tackle the problem of learning from biased data in the very realistic unsupervised scenario, i.e., when the bias is unknown. This is a much harder task as compared to the supervised case, where auxiliary, bias-related annotations, can be exploited in the learning process. This paper proposes a novel 2-stage learning pipeline featuring a data augmentation strategy able to regularize the training. First, biased/unbiased samples are identified by training over-biased models. Second, such subdivision (typically noisy) is exploited within a data augmentation framework, properly combining the original samples while learning mixing parameters, which has a regularization effect. Experiments on synthetic and realistic biased datasets show state-of-the-art classification accuracy, outperforming competing methods, ultimately proving robust performance on both biased and unbiased examples. Notably, being our training method totally agnostic to the level of bias, it also positively affects performance for any, even apparently unbiased, dataset, thus improving the model generalization regardless of the level of bias (or its absence) in the data.