- Published 2013 in ICML

Deep and recurrent neural networks (DNNs and RNNs respectively) are powerful models that were considered to be almost impossible to train using stochastic gradient descent with momentum. In this paper, we show that when stochastic gradient descent with momentum uses a well-designed random initialization and a particular type of slowly increasing schedule for the momentum parameter, it can train both DNNs and RNNs (on datasets with long-term dependencies) to levels of performance that were previously achievable only with Hessian-Free optimization. We find that both the initialization and the momentum are crucial since poorly initialized networks cannot be trained with momentum and well-initialized networks perform markedly worse when the momentum is absent or poorly tuned. Our success training these models suggests that previous attempts to train deep and recurrent neural networks from random initializations have likely failed due to poor initialization schemes. Furthermore, carefully tuned momentum methods suffice for dealing with the curvature issues in deep and recurrent network training objectives without the need for sophisticated second-order methods.

Citations per Year

Semantic Scholar estimates that this publication has **650** citations based on the available data.

See our **FAQ** for additional information.

Showing 1-10 of 446 extracted citations

Highly Influenced

4 Excerpts

Highly Influenced

4 Excerpts

Highly Influenced

5 Excerpts

Highly Influenced

12 Excerpts

Highly Influenced

4 Excerpts

Highly Influenced

7 Excerpts

Highly Influenced

5 Excerpts

Highly Influenced

4 Excerpts

Highly Influenced

8 Excerpts

Highly Influenced

6 Excerpts

@inproceedings{Sutskever2013OnTI,
title={On the importance of initialization and momentum in deep learning},
author={Ilya Sutskever and James Martens and George E. Dahl and Geoffrey E. Hinton},
booktitle={ICML},
year={2013}
}