Relationship between Model Compression and Adversarial Robustness: A Review of Current Evidence
It addresses the problem of understanding how compression affects robustness for researchers and practitioners in machine learning, but it is incremental as it reviews existing studies without new experiments.
This review paper examines the relationship between model compression techniques, such as pruning and quantization, and adversarial robustness in deep learning, summarizing current evidence and discussing contradictory findings.
Increasing the model capacity is a known approach to enhance the adversarial robustness of deep learning networks. On the other hand, various model compression techniques, including pruning and quantization, can reduce the size of the network while preserving its accuracy. Several recent studies have addressed the relationship between model compression and adversarial robustness, while some experiments have reported contradictory results. This work summarizes available evidence and discusses possible explanations for the observed effects.