Training Adversarially Robust Sparse Networks via Bayesian Connectivity Sampling

Ozan Özdenizci, Robert Legenstein
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:8314-8324, 2021.

Abstract

Deep neural networks have been shown to be susceptible to adversarial attacks. This lack of adversarial robustness is even more pronounced when models are compressed in order to meet hardware limitations. Hence, if adversarial robustness is an issue, training of sparsely connected networks necessitates considering adversarially robust sparse learning. Motivated by the efficient and stable computational function of the brain in the presence of a highly dynamic synaptic connectivity structure, we propose an intrinsically sparse rewiring approach to train neural networks with state-of-the-art robust learning objectives under high sparsity. Importantly, in contrast to previously proposed pruning techniques, our approach satisfies global connectivity constraints throughout robust optimization, i.e., it does not require dense pre-training followed by pruning. Based on a Bayesian posterior sampling principle, a network rewiring process simultaneously learns the sparse connectivity structure and the robustness-accuracy trade-off based on the adversarial learning objective. Although our networks are sparsely connected throughout the whole training process, our experimental benchmark evaluations show that their performance is superior to recently proposed robustness-aware network pruning methods which start from densely connected networks.

Cite this Paper


BibTeX
@InProceedings{pmlr-v139-ozdenizci21a, title = {Training Adversarially Robust Sparse Networks via Bayesian Connectivity Sampling}, author = {{\"O}zdenizci, Ozan and Legenstein, Robert}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {8314--8324}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/ozdenizci21a/ozdenizci21a.pdf}, url = {https://proceedings.mlr.press/v139/ozdenizci21a.html}, abstract = {Deep neural networks have been shown to be susceptible to adversarial attacks. This lack of adversarial robustness is even more pronounced when models are compressed in order to meet hardware limitations. Hence, if adversarial robustness is an issue, training of sparsely connected networks necessitates considering adversarially robust sparse learning. Motivated by the efficient and stable computational function of the brain in the presence of a highly dynamic synaptic connectivity structure, we propose an intrinsically sparse rewiring approach to train neural networks with state-of-the-art robust learning objectives under high sparsity. Importantly, in contrast to previously proposed pruning techniques, our approach satisfies global connectivity constraints throughout robust optimization, i.e., it does not require dense pre-training followed by pruning. Based on a Bayesian posterior sampling principle, a network rewiring process simultaneously learns the sparse connectivity structure and the robustness-accuracy trade-off based on the adversarial learning objective. Although our networks are sparsely connected throughout the whole training process, our experimental benchmark evaluations show that their performance is superior to recently proposed robustness-aware network pruning methods which start from densely connected networks.} }
Endnote
%0 Conference Paper %T Training Adversarially Robust Sparse Networks via Bayesian Connectivity Sampling %A Ozan Özdenizci %A Robert Legenstein %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-ozdenizci21a %I PMLR %P 8314--8324 %U https://proceedings.mlr.press/v139/ozdenizci21a.html %V 139 %X Deep neural networks have been shown to be susceptible to adversarial attacks. This lack of adversarial robustness is even more pronounced when models are compressed in order to meet hardware limitations. Hence, if adversarial robustness is an issue, training of sparsely connected networks necessitates considering adversarially robust sparse learning. Motivated by the efficient and stable computational function of the brain in the presence of a highly dynamic synaptic connectivity structure, we propose an intrinsically sparse rewiring approach to train neural networks with state-of-the-art robust learning objectives under high sparsity. Importantly, in contrast to previously proposed pruning techniques, our approach satisfies global connectivity constraints throughout robust optimization, i.e., it does not require dense pre-training followed by pruning. Based on a Bayesian posterior sampling principle, a network rewiring process simultaneously learns the sparse connectivity structure and the robustness-accuracy trade-off based on the adversarial learning objective. Although our networks are sparsely connected throughout the whole training process, our experimental benchmark evaluations show that their performance is superior to recently proposed robustness-aware network pruning methods which start from densely connected networks.
APA
Özdenizci, O. & Legenstein, R.. (2021). Training Adversarially Robust Sparse Networks via Bayesian Connectivity Sampling. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:8314-8324 Available from https://proceedings.mlr.press/v139/ozdenizci21a.html.

Related Material