A Line-Search Descent Algorithm for Strict Saddle Functions with Guaranteed Complexity
Authors: Michael J. O’Neill, Stephen J. Wright; Published: 2023, Volume 24, Issue 10, Pages 1-34.
Abstract
This article presents a line-search algorithm that achieves the best-known worst-case complexity results for problems with a specific property known as the “strict saddle” property. This property has been observed to exist in low-rank matrix optimization problems. The algorithm is adaptive, utilizing backtracking line searches and eliminating the need for prior knowledge of the parameters defining the strict saddle property.
[Abstract]