The computational complexity of finding stationary points in non-convex optimization

期刊论文
出版模式：

开放获取
发表日期：

2024年9月27日

查看全文

Alexandros Hollender,
Manolis Zampetakis

摘要

Finding approximate stationary points, i.e., points where the gradient is approximately zero, of non-convex but smooth objective functions f over unrestricted d-dimensional domains is one of the most fundamental problems in classical non-convex optimization. Nevertheless, the computational and query complexity of this problem are still not well understood when the dimension d of the problem is independent of the approximation error. In this paper, we show the following computational and query complexity results: 1.

The problem of finding approximate stationary points over unrestricted domains is ${\textsf{PLS}} $ $PLS$ -complete.

For $d = 2$ $d = 2$ , we provide a zero-order algorithm for finding $\varepsilon $ $ε$ -approximate stationary points that requires at most $O(1/\varepsilon )$ $O (1 / ε)$ value queries to the objective function.

We show that any algorithm needs at least $\Omega (1/\varepsilon )$ $Ω (1 / ε)$ queries to the objective function and/or its gradient to find $\varepsilon $ $ε$ -approximate stationary points when $d=2$ $d = 2$ . Combined with the above, this characterizes the query complexity of this problem to be $\Theta (1/\varepsilon )$ $Θ (1 / ε)$ .

For $d = 2$ $d = 2$ , we provide a zero-order algorithm for finding $\varepsilon $ $ε$ -KKT points in constrained optimization problems that requires at most $O(1/\sqrt{\varepsilon })$ $O (1 / \sqrt{ε})$ value queries to the objective function. This closes the gap between works of Bubeck and Mikulincer and Vavasis and characterizes the query complexity of this problem to be $\Theta (1/\sqrt{\varepsilon })$ $Θ (1 / \sqrt{ε})$ .

Combining our results with a recent result of Fearnley et al., we show that finding approximate KKT points in constrained optimization is reducible to finding approximate stationary points in unconstrained optimization but the converse is impossible.