Timezone: »

 
Poster
Model Compression with Adversarial Robustness: A Unified Optimization Framework
Shupeng Gui · Haotao Wang · Haichuan Yang · Chen Yu · Zhangyang Wang · Ji Liu

Tue Dec 10 10:45 AM -- 12:45 PM (PST) @ East Exhibition Hall B + C #116
in Deep Learning -- Efficient Inference Methods »

Deep model compression has been extensively studied, and state-of-the-art methods can now achieve high compression ratios with minimal accuracy loss. This paper studies model compression through a different lens: could we compress models without hurting their robustness to adversarial attacks, in addition to maintaining accuracy? Previous literature suggested that the goals of robustness and compactness might sometimes contradict. We propose a novel Adversarially Trained Model Compression (ATMC) framework. ATMC constructs a unified constrained optimization formulation, where existing compression means (pruning, factorization, quantization) are all integrated into the constraints. An efficient algorithm is then developed. An extensive group of experiments are presented, demonstrating that ATMC obtains remarkably more favorable trade-off among model size, accuracy and robustness, over currently available alternatives in various settings. The codes are publicly available at: https://github.com/shupenggui/ATMC.

Keywords: Deep Learning -> Efficient Inference Methods · Algorithms -> Adversarial Learning

Author Information

Shupeng Gui (University of Rochester)
Haotao Wang (Texas A&M University)
Haichuan Yang (University of Rochester)
Chen Yu (University of Rochester)
Zhangyang Wang (TAMU)
Ji Liu (Kwai Inc.)

More from the Same Authors