Neural Population Geometry Reveals the Role of Stochasticity in Robust Perception

Joel Dapello, Jenelle Feather, Hang Le, Tiago Marques, David Daniel Cox, Josh Mcdermott, James J. DiCarlo, SueYeon Chung

Keywords: Representational Geometry, Biologically inspired vision models, Biologically inspired auditory models, Stochasticity, Adversarial robustness, Robust perception

TL;DR: Adversarially trained networks and biologically inspired stochastic networks in both visual and auditory domains demonstrate distinct mechanisms for robust perception as revealed by neural population geometry.

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 demonstrate 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.

Code Of Conduct: I certify that all co-authors of this work have read and commit to adhering to the NeurIPS Statement on Ethics, Fairness, Inclusivity, and Code of Conduct.

Code: https://github.com/chung-neuroai-lab/adversarial-manifolds