Multicoated Supermasks Enhance Hidden Networks
Yasuyuki Okoshi, Ángel López Garcı́a-Arias, Kazutoshi Hirose, Kota Ando, Kazushi Kawamura, Thiem Van Chu, Masato Motomura, Jaehoon Yu
Abstract
Hidden Networks (Ramanujan et al., 2020) showed the possibility of finding accurate subnetworks within a randomly weighted neural network by training a connectivity mask, referred to as supermask. We show that the supermask stops improving even though gradients are not zero, thus underutilizing backpropagated information. To address this we propose a method that extends Hidden Networks by training an overlay of multiple hierarchical supermasks{—}a multicoated supermask. This method shows that using multiple supermasks for a single task achieves higher accuracy without additional training cost. Experiments on CIFAR-10 and ImageNet show that Multicoated Supermasks enhance the tradeoff between accuracy and model size. A ResNet-101 using a 7-coated supermask outperforms its Hidden Networks counterpart by 4%, matching the accuracy of a dense ResNet-50 while being an order of magnitude smaller.
BibTeX
@InProceedings{pmlr-v162-okoshi22a,
title = {Multicoated Supermasks Enhance Hidden Networks},
author = {Okoshi, Yasuyuki and Garc\'{\i}a-Arias, \'Angel L{\'o}pez and Hirose, Kazutoshi and Ando, Kota and Kawamura, Kazushi and Van Chu, Thiem and Motomura, Masato and Yu, Jaehoon},
booktitle = {Proceedings of the 39th International Conference on Machine Learning},
pages = {17045--17055},
year = {2022},
editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan},
volume = {162},
series = {Proceedings of Machine Learning Research},
month = {17--23 Jul},
publisher = {PMLR},
pdf = {https://proceedings.mlr.press/v162/okoshi22a/okoshi22a.pdf},
url = {https://proceedings.mlr.press/v162/okoshi22a.html},
abstract = {Hidden Networks (Ramanujan et al., 2020) showed the possibility of finding accurate subnetworks within a randomly weighted neural network by training a connectivity mask, referred to as supermask. We show that the supermask stops improving even though gradients are not zero, thus underutilizing backpropagated information. To address this we propose a method that extends Hidden Networks by training an overlay of multiple hierarchical supermasks{—}a multicoated supermask. This method shows that using multiple supermasks for a single task achieves higher accuracy without additional training cost. Experiments on CIFAR-10 and ImageNet show that Multicoated Supermasks enhance the tradeoff between accuracy and model size. A ResNet-101 using a 7-coated supermask outperforms its Hidden Networks counterpart by 4%, matching the accuracy of a dense ResNet-50 while being an order of magnitude smaller.}
}