Improving Robustness of Jet Tagging Algorithms with Adversarial Training

TL;DR

This paper introduces an adversarial training approach to enhance the robustness of jet flavor tagging algorithms in high-energy physics, addressing vulnerabilities caused by simulation mismodeling.

Contribution

The study develops and demonstrates an adversarial training method that reduces the impact of mismodeling on neural network-based jet tagging algorithms.

Findings

Adversarial training improves classifier robustness against mismodeling.

The method reduces performance degradation due to simulation inaccuracies.

Enhanced robustness leads to more reliable physics object identification.

Abstract

Deep learning is a standard tool in the field of high-energy physics, facilitating considerable sensitivity enhancements for numerous analysis strategies. In particular, in identification of physics objects, such as jet flavor tagging, complex neural network architectures play a major role. However, these methods are reliant on accurate simulations. Mismodeling can lead to non-negligible differences in performance in data that need to be measured and calibrated against. We investigate the classifier response to input data with injected mismodelings and probe the vulnerability of flavor tagging algorithms via application of adversarial attacks. Subsequently, we present an adversarial training strategy that mitigates the impact of such simulated attacks and improves the classifier robustness. We examine the relationship between performance and vulnerability and show that this method constitutes a promising approach to reduce the vulnerability to poor modeling.