ODE guided Neural Data Augmentation Techniques for Time Series Data and its Benefits on Robustness

TL;DR

This paper introduces ODE-guided neural data augmentation methods for time series that enhance classification accuracy and robustness against adversarial attacks and corruptions, addressing a gap in deep learning for time series data.

Contribution

It proposes two local gradient-based and one spectral density-based data augmentation techniques, demonstrating their effectiveness in improving accuracy and robustness of time series classifiers.

Findings

Achieved state-of-the-art accuracy on multiple benchmarks.

Enhanced robustness against FGSM and BIM attacks.

Improved model performance with novel augmentation methods.

Abstract

Exploring adversarial attack vectors and studying their effects on machine learning algorithms has been of interest to researchers. Deep neural networks working with time series data have received lesser interest compared to their image counterparts in this context. In a recent finding, it has been revealed that current state-of-the-art deep learning time series classifiers are vulnerable to adversarial attacks. In this paper, we introduce two local gradient based and one spectral density based time series data augmentation techniques. We show that a model trained with data obtained using our techniques obtains state-of-the-art classification accuracy on various time series benchmarks. In addition, it improves the robustness of the model against some of the most common corruption techniques,such as Fast Gradient Sign Method (FGSM) and Basic Iterative Method (BIM).