Neural Population Geometry Reveals the Role of Stochasticity in Robust Perception

TL;DR

This paper investigates how neural response stochasticity influences the internal representations of neural networks, revealing geometric signatures that relate to robustness against adversarial attacks in both visual and auditory domains.

Contribution

It introduces a geometric analysis of stochastic neural networks, demonstrating their role in enhancing adversarial robustness and uncovering mechanisms behind robust perception.

Findings

Distinct geometric signatures for different network types

Neural stochasticity overlaps representations of clean and adversarial stimuli

Tradeoff between adversarial robustness and clean performance

Abstract

Adversarial examples are often cited by neuroscientists and machine learning researchers as an example of how computational models diverge from biological sensory systems. Recent work has proposed adding biologically-inspired components to visual neural networks as a way to improve their adversarial robustness. One surprisingly effective component for reducing adversarial vulnerability is response stochasticity, like that exhibited by biological neurons. Here, using recently developed geometrical techniques from computational neuroscience, we investigate how adversarial perturbations influence the internal representations of standard, adversarially trained, and biologically-inspired stochastic networks. We find distinct geometric signatures for each type of network, revealing different mechanisms for achieving robust representations. Next, we generalize these results to the auditory domain, showing that neural stochasticity also makes auditory models more robust to adversarial perturbations. Geometric analysis of the stochastic networks reveals overlap between representations of clean and adversarially perturbed stimuli, and quantitatively demonstrates that competing geometric effects of stochasticity mediate a tradeoff between adversarial and clean performance. Our results shed light on the strategies of robust perception utilized by adversarially trained and stochastic networks, and help explain how stochasticity may be beneficial to machine and biological computation.