Skip to Main content Skip to Navigation
Conference papers

A Two-Stage Subspace Trust Region Approach for Deep Neural Network Training

Abstract : In this paper, we develop a novel second-order method for training feed-forward neural nets. At each iteration, we construct a quadratic approximation to the cost function in a low-dimensional subspace. We minimize this approximation inside a trust region through a two-stage procedure: first inside the embedded positive curvature subspace, followed by a gradient descent step. This approach leads to a fast objective function decay, prevents convergence to saddle points, and alleviates the need for manually tuning parameters. We show the good performance of the proposed algorithm on benchmark datasets.
Complete list of metadatas

Cited literature [22 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01634538
Contributor : Emilie Chouzenoux <>
Submitted on : Tuesday, November 14, 2017 - 11:32:35 AM
Last modification on : Wednesday, April 8, 2020 - 3:27:10 PM
Document(s) archivé(s) le : Thursday, February 15, 2018 - 2:50:52 PM

File

PID4857015.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01634538, version 1

Citation

Viacheslav Dudar, Giovanni Chierchia, Emilie Chouzenoux, Jean-Christophe Pesquet, Vladimir Semenov. A Two-Stage Subspace Trust Region Approach for Deep Neural Network Training. 25th European Signal Processing Conference (EUSIPCO 2017), Aug 2017, Kos Island, Greece. pp.291-295. ⟨hal-01634538⟩

Share

Metrics

Record views

687

Files downloads

370