Achieving Adversarial Robustness via Sparsity

Abstract

Network pruning has been known to produce compact models without much accuracy degradation. However, how the pruning process affects a network's robustness and the working mechanism behind remain unresolved. In this work, we theoretically prove that the sparsity of network weights is closely associated with model robustness. Through experiments on a variety of adversarial pruning methods, we find that weights sparsity will not hurt but improve robustness, where both weights inheritance from the lottery ticket and adversarial training improve model robustness in network pruning. Based on these findings, we propose a novel adversarial training method called inverse weights inheritance, which imposes sparse weights distribution on a large network by inheriting weights from a small network, thereby improving the robustness of the large network.

References in corpus (29)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Explaining and Harnessing Adversarial Examples
• Towards Deep Learning Models Resistant to Adversarial Attacks
• The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks
• Rethinking the Value of Network Pruning
• Theoretically Principled Trade-off between Robustness and Accuracy
• Pruning Filters for Efficient ConvNets
• On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima
• Adversarial Machine Learning at Scale
• Formal Guarantees on the Robustness of a Classifier against Adversarial Manipulation
• Adversarial Training Methods for Semi-Supervised Text Classification
• Stochastic Activation Pruning for Robust Adversarial Defense
• Adversarially Robust Distillation
• Rethinking the Smaller-Norm-Less-Informative Assumption in Channel Pruning of Convolution Layers
• Robustness of classifiers: from adversarial to random noise
• One ticket to win them all: generalizing lottery ticket initializations across datasets and optimizers
• Sparse DNNs with Improved Adversarial Robustness
• Stronger generalization bounds for deep nets via a compression approach
• Interpreting Adversarially Trained Convolutional Neural Networks
• ME-Net: Towards Effective Adversarial Robustness with Matrix Estimation
• Model Compression with Adversarial Robustness: A Unified Optimization Framework
• The Generalization-Stability Tradeoff In Neural Network Pruning
• Towards Compact and Robust Deep Neural Networks
• Adversarial Neural Pruning with Latent Vulnerability Suppression
• To compress or not to compress: Understanding the Interactions between Adversarial Attacks and Neural Network Compression
• Robust Sparse Regularization: Simultaneously Optimizing Neural Network Robustness and Compactness
• The Search for Sparse, Robust Neural Networks
• ResNets Ensemble via the Feynman-Kac Formalism to Improve Natural and Robust Accuracies
• Adversarial Risk Bounds for Neural Networks through Sparsity based Compression

Cited by in corpus (3)

• Neural network relief: a pruning algorithm based on neural activity
• Adversarial Pruning: A Survey and Benchmark of Pruning Methods for Adversarial Robustness
• A Winning Hand: Compressing Deep Networks Can Improve Out-Of-Distribution Robustness