Back to Search Start Over

First-order methods almost always avoid strict saddle points

Authors :
Georgios Piliouras
Max Simchowitz
Benjamin Recht
Michael I. Jordan
Ioannis Panageas
Jason D. Lee
Source :
Mathematical Programming. 176:311-337
Publication Year :
2019
Publisher :
Springer Science and Business Media LLC, 2019.

Abstract

We establish that first-order methods avoid strict saddle points for almost all initializations. Our results apply to a wide variety of first-order methods, including (manifold) gradient descent, block coordinate descent, mirror descent and variants thereof. The connecting thread is that such algorithms can be studied from a dynamical systems perspective in which appropriate instantiations of the Stable Manifold Theorem allow for a global stability analysis. Thus, neither access to second-order derivative information nor randomness beyond initialization is necessary to provably avoid strict saddle points.

Details

ISSN :
14364646 and 00255610
Volume :
176
Database :
OpenAIRE
Journal :
Mathematical Programming
Accession number :
edsair.doi...........c09d33eb51ea28786e4d899aae590b56
Full Text :
https://doi.org/10.1007/s10107-019-01374-3