Neural Population Geometry Reveals the Role of Stochasticity in Robust Perception
This work addresses adversarial vulnerability in machine learning models by exploring biologically-inspired stochasticity, offering insights for robust perception in both artificial and biological systems, though it is incremental as it builds on existing geometric techniques and stochastic components.
The study investigated how adversarial perturbations affect internal representations in neural networks, finding distinct geometric signatures for standard, adversarially trained, and biologically-inspired stochastic networks, with stochasticity improving robustness in both visual and auditory domains while mediating a tradeoff between adversarial and clean performance.
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.