ASAT: Adaptively Scaled Adversarial Training in Time Series

TL;DR

This paper introduces ASAT, an adaptive adversarial training method for time series neural networks, which improves generalization and robustness by rescaling data at different time slots, demonstrated in finance applications.

Contribution

The paper proposes a novel adaptive scaling approach for adversarial training in time series analysis, enhancing neural network robustness and generalization capabilities.

Findings

ASAT improves neural network robustness against adversarial attacks.

ASAT enhances generalization ability compared to traditional adversarial training.

Experimental results in finance demonstrate the effectiveness of ASAT.

Abstract

Adversarial training is a method for enhancing neural networks to improve the robustness against adversarial examples. Besides the security concerns of potential adversarial examples, adversarial training can also improve the generalization ability of neural networks, train robust neural networks, and provide interpretability for neural networks. In this work, we introduce adversarial training in time series analysis to enhance the neural networks for better generalization ability by taking the finance field as an example. Rethinking existing research on adversarial training, we propose the adaptively scaled adversarial training (ASAT) in time series analysis, by rescaling data at different time slots with adaptive scales. Experimental results show that the proposed ASAT can improve both the generalization ability and the adversarial robustness of neural networks compared to the baselines. Compared to the traditional adversarial training algorithm, ASAT can achieve better generalization ability and similar adversarial robustness.