Showing 1–1 of 1 results for author: Matet, S
-
Don't relax: early stop** for convex regularization
Authors:
Simon Matet,
Lorenzo Rosasco,
Silvia Villa,
Bang Long Vu
Abstract:
We consider the problem of designing efficient regularization algorithms when regularization is encoded by a (strongly) convex functional. Unlike classical penalization methods based on a relaxation approach, we propose an iterative method where regularization is achieved via early stop**. Our results show that the proposed procedure achieves the same recovery accuracy as penalization methods, w…
▽ More
We consider the problem of designing efficient regularization algorithms when regularization is encoded by a (strongly) convex functional. Unlike classical penalization methods based on a relaxation approach, we propose an iterative method where regularization is achieved via early stop**. Our results show that the proposed procedure achieves the same recovery accuracy as penalization methods, while naturally integrating computational considerations. An empirical analysis on a number of problems provides promising results with respect to the state of the art.
△ Less
Submitted 17 July, 2017;
originally announced July 2017.
