Skip to main content
Log in
(current)
MIT Open Scholarship
DSpace@MIT
Search
Home
MIT Open Access Articles
MIT Open Access Articles
Complexity control by gradient descent in deep networks
This is not the latest version of this item. The latest version can be found
here
.
Complexity control by gradient descent in deep networks
Download
Name
s41467-020-14663-9.pdf
Description
Published version
Show more
Size
431.68 KB
Format
Adobe PDF
Checksum
(MD5)
cb2bfa9216379d5f6becece6d420b4e7
Show more
sword-2021-03-24T13:05:49.original.xml (130 B)
Original SWORD entry document
Show more
Author(s)
Poggio, Tomaso
•
Liao, Qianli
•
Banburski, Andrzej
Date Issued
2020
Journal
Nature Communications
Publisher
Springer Science and Business Media LLC
Version
Final published version
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.
Show more
Terms of Use
Creative Commons Attribution 4.0 International license
https://creativecommons.org/licenses/by/4.0/
Persistent DSpace Link
https://hdl.handle.net/1721.1/136301
DOI of Published Version
10.1038/S41467-020-14663-9
