Deterministic versus stochastic dynamical classifiers: opposing random adversarial attacks with noise

TL;DR

This paper compares deterministic and stochastic dynamical classifiers, demonstrating that the stochastic CVFR model is robust against adversarial noise, highlighting the beneficial interplay of noise and dynamics in classification tasks.

Contribution

It introduces a stochastic variant of the CVFR model as a classifier and shows its robustness to adversarial noise, revealing new insights into noise-dynamics synergy.

Findings

Stochastic CVFR is resilient to adversarial attacks.

Dynamical classifiers can leverage noise for robustness.

Spectral embedding of attractors guides classification.

Abstract

The Continuous-Variable Firing Rate (CVFR) model, widely used in neuroscience to describe the intertangled dynamics of excitatory biological neurons, is here trained and tested as a veritable dynamically assisted classifier. To this end the model is supplied with a set of planted attractors which are self-consistently embedded in the inter-nodes coupling matrix, via its spectral decomposition. Learning to classify amounts to sculp the basin of attraction of the imposed equilibria, directing different items towards the corresponding destination target, which reflects the class of respective pertinence. A stochastic variant of the CVFR model is also studied and found to be robust to aversarial random attacks, which corrupt the items to be classified. This remarkable finding is one of the very many surprising effects which arise when noise and dynamical attributes are made to mutually resonate.