Regularizing deep networks using efficient layerwise adversarial training

TL;DR

This paper introduces an efficient layerwise adversarial training method that perturbs intermediate activations to regularize deep networks, improving robustness and accuracy on CIFAR datasets.

Contribution

The paper proposes a novel layerwise adversarial training approach that enhances deep network regularization and performance, outperforming existing methods like dropout.

Findings

Improved accuracy on CIFAR-10 and CIFAR-100 datasets.

Enhanced robustness to adversarial examples.

Outperforms dropout and larger models on WideResNets.

Abstract

Adversarial training has been shown to regularize deep neural networks in addition to increasing their robustness to adversarial examples. However, its impact on very deep state of the art networks has not been fully investigated. In this paper, we present an efficient approach to perform adversarial training by perturbing intermediate layer activations and study the use of such perturbations as a regularizer during training. We use these perturbations to train very deep models such as ResNets and show improvement in performance both on adversarial and original test data. Our experiments highlight the benefits of perturbing intermediate layer activations compared to perturbing only the inputs. The results on CIFAR-10 and CIFAR-100 datasets show the merits of the proposed adversarial training approach. Additional results on WideResNets show that our approach provides significant improvement in classification accuracy for a given base model, outperforming dropout and other base models of larger size.