Large-Scale Dataset Pruning in Adversarial Training through Data Importance Extrapolation

TL;DR

This paper introduces a data pruning method for adversarial training that extrapolates data importance from small samples to large datasets, reducing training data without sacrificing robustness.

Contribution

It presents a novel data importance extrapolation technique specifically designed for adversarial training to efficiently prune datasets.

Findings

Reduces dataset size while maintaining robustness

Efficiently extrapolates importance scores from small to large datasets

Improves training efficiency in adversarial settings

Abstract

Their vulnerability to small, imperceptible attacks limits the adoption of deep learning models to real-world systems. Adversarial training has proven to be one of the most promising strategies against these attacks, at the expense of a substantial increase in training time. With the ongoing trend of integrating large-scale synthetic data this is only expected to increase even further. Thus, the need for data-centric approaches that reduce the number of training samples while maintaining accuracy and robustness arises. While data pruning and active learning are prominent research topics in deep learning, they are as of now largely unexplored in the adversarial training literature. We address this gap and propose a new data pruning strategy based on extrapolating data importance scores from a small set of data to a larger set. In an empirical evaluation, we demonstrate that extrapolation-based pruning can efficiently reduce dataset size while maintaining robustness.