Current browse context:
math.OC
Change to browse by:
References & Citations
Mathematics > Optimization and Control
Title: A Stochastic Trust Region Algorithm Based on Careful Step Normalization
(Submitted on 29 Dec 2017 (v1), last revised 26 Jun 2018 (this version, v3))
Abstract: An algorithm is proposed for solving stochastic and finite sum minimization problems. Based on a trust region methodology, the algorithm employs normalized steps, at least as long as the norms of the stochastic gradient estimates are within a specified interval. The complete algorithm---which dynamically chooses whether or not to employ normalized steps---is proved to have convergence guarantees that are similar to those possessed by a traditional stochastic gradient approach under various sets of conditions related to the accuracy of the stochastic gradient estimates and choice of stepsize sequence. The results of numerical experiments are presented when the method is employed to minimize convex and nonconvex machine learning test problems. These results illustrate that the method can outperform a traditional stochastic gradient approach.
Submission history
From: Frank E. Curtis [view email][v1] Fri, 29 Dec 2017 16:59:25 GMT (447kb,D)
[v2] Sat, 27 Jan 2018 16:19:08 GMT (447kb,D)
[v3] Tue, 26 Jun 2018 16:14:56 GMT (624kb,D)
Link back to: arXiv, form interface, contact.