Fermi-Bose Machine achieves both generalization and adversarial robustness

TL;DR

This paper introduces a biologically plausible local contrastive learning method inspired by Fermi-Bose statistics, which enhances generalization and significantly reduces adversarial vulnerability in neural networks.

Contribution

It proposes a novel local contrastive learning approach that replaces backpropagation, inspired by Fermi-Bose statistics, to improve robustness and generalization.

Findings

Reduces adversarial vulnerability on MNIST

Controls geometric separation of prototypes

Achieves better generalization without backpropagation

Abstract

Distinct from human cognitive processing, deep neural networks trained by backpropagation can be easily fooled by adversarial examples. To design a semantically meaningful representation learning, we discard backpropagation, and instead, propose a local contrastive learning, where the representation for the inputs bearing the same label shrink (akin to boson) in hidden layers, while those of different labels repel (akin to fermion). This layer-wise learning is local in nature, being biological plausible. A statistical mechanics analysis shows that the target fermion-pair-distance is a key parameter. Moreover, the application of this local contrastive learning to MNIST benchmark dataset demonstrates that the adversarial vulnerability of standard perceptron can be greatly mitigated by tuning the target distance, i.e., controlling the geometric separation of prototype manifolds.