Notice

This is not the latest version of this item. The latest version can be found at:https://dspace.mit.edu/handle/1721.1/136301.2

Complexity control by gradient descent in deep networks

Author(s)
Poggio, Tomaso; Liao, Qianli; Banburski, Andrzej
Thumbnail
DownloadPublished version (431.6Kb)
Publisher with Creative Commons License
Terms of use
Creative Commons Attribution 4.0 International license https://creativecommons.org/licenses/by/4.0/
Metadata
Show full item record
Abstract
© 2020, The Author(s). Overparametrized deep networks predict well, despite the lack of an explicit complexity control during training, such as an explicit regularization term. For exponential-type loss functions, we solve this puzzle by showing an effective regularization effect of gradient descent in terms of the normalized weights that are relevant for classification.
Date issued
2020
URI
https://hdl.handle.net/1721.1/136301
Journal
Nature Communications
Publisher
Springer Science and Business Media LLC
Collections

Version History

VersionItemDateSummary
21721.1/136301.22022-01-12T20:48:30ZAuthority information verified/added.
11721.1/136301*2021-10-27T20:34:47Z

*Selected version