Noisy Learning for Neural ODEs Acts as a Robustness Locus Widening

TL;DR

This paper introduces a new metric and methodology for evaluating the robustness of neural ODEs against data corruptions, demonstrating that simple data augmentation can enhance their natural robustness.

Contribution

It proposes a novel accuracy metric and evaluation methodology for neural ODE robustness, and shows data augmentation improves their resilience to image corruptions.

Findings

Neural ODEs exhibit natural robustness to image corruptions.

The proposed metric effectively evaluates robustness and dataset simulators.

Data augmentation enhances neural ODEs' robustness across datasets.

Abstract

We investigate the problems and challenges of evaluating the robustness of Differential Equation-based (DE) networks against synthetic distribution shifts. We propose a novel and simple accuracy metric which can be used to evaluate intrinsic robustness and to validate dataset corruption simulators. We also propose methodology recommendations, destined for evaluating the many faces of neural DEs' robustness and for comparing them with their discrete counterparts rigorously. We then use this criteria to evaluate a cheap data augmentation technique as a reliable way for demonstrating the natural robustness of neural ODEs against simulated image corruptions across multiple datasets.