Fermi-Bose Machine achieves both generalization and adversarial robustness
This addresses the issue of adversarial attacks in machine learning, offering a biologically plausible approach, but it appears incremental as it builds on existing contrastive learning ideas with specific parameter tuning.
The paper tackles the problem of adversarial vulnerability in deep neural networks by proposing a local contrastive learning method that replaces backpropagation, achieving improved adversarial robustness on the MNIST benchmark dataset through tuning geometric separation parameters.
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.