This is not the latest version of this item. The latest version can be found here.
Are deep ResNets provably better than linear predictors?
Name
NeurIPS-2019-are-deep-resnets-provably-better-than-linear-predictors-Paper.pdf
Description
Published version
Size
284.26 KB
Format
Adobe PDF
Checksum (MD5)
7016ce26fe567abc92947ecaf326b681
Date Issued
December 2019
Journal
Advances in Neural Information Processing Systems
Citation
2019. "Are deep ResNets provably better than linear predictors?." Advances in Neural Information Processing Systems, 32.
Version
Final published version
Abstract
© 2019 Neural information processing systems foundation. All rights reserved. Recent results in the literature indicate that a residual network (ResNet) composed of a single residual block outperforms linear predictors, in the sense that all local minima in its optimization landscape are at least as good as the best linear predictor. However, these results are limited to a single residual block (i.e., shallow ResNets), instead of the deep ResNets composed of multiple residual blocks. We take a step towards extending this result to deep ResNets. We start by two motivating examples. First, we show that there exist datasets for which all local minima of a fully-connected ReLU network are no better than the best linear predictor, whereas a ResNet has strictly better local minima. Second, we show that even at the global minimum, the representation obtained from the residual block outputs of a 2-block ResNet do not necessarily improve monotonically over subsequent blocks, which highlights a fundamental difficulty in analyzing deep ResNets. Our main theorem on deep ResNets shows under simple geometric conditions that, any critical point in the optimization landscape is either (i) at least as good as the best linear predictor; or (ii) the Hessian at this critical point has a strictly negative eigenvalue. Notably, our theorem shows that a chain of multiple skip-connections can improve the optimization landscape, whereas existing results study direct skip-connections to the last hidden layer or output layer. Finally, we complement our results by showing benign properties of the “near-identity regions” of deep ResNets, showing depth-independent upper bounds for the risk attained at critical points as well as the Rademacher complexity.
Terms of Use
Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use.
Persistent DSpace Link
DOI of Published Version
https://papers.nips.cc/paper/2019/hash/661c1c090ff5831a647202397c61d73c-Abstract.html
