Curvature-Aware Derivative-Free Optimization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kim, Bumsu, Cai, HanQin, McKenzie, Daniel, Yin, Wotao
Format: Preprint
Published: 2021
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912329996697600
author Kim, Bumsu
Cai, HanQin
McKenzie, Daniel
Yin, Wotao
author_facet Kim, Bumsu
Cai, HanQin
McKenzie, Daniel
Yin, Wotao
contents The paper discusses derivative-free optimization (DFO), which involves minimizing a function without access to gradients or directional derivatives, only function evaluations. Classical DFO methods, which mimic gradient-based methods, such as Nelder-Mead and direct search have limited scalability for high-dimensional problems. Zeroth-order methods have been gaining popularity due to the demands of large-scale machine learning applications, and the paper focuses on the selection of the step size $α_k$ in these methods. The proposed approach, called Curvature-Aware Random Search (CARS), uses first- and second-order finite difference approximations to compute a candidate $α_{+}$. We prove that for strongly convex objective functions, CARS converges linearly provided that the search direction is drawn from a distribution satisfying very mild conditions. We also present a Cubic Regularized variant of CARS, named CARS-CR, which converges in a rate of $\mathcal{O}(k^{-1})$ without the assumption of strong convexity. Numerical experiments show that CARS and CARS-CR match or exceed the state-of-the-arts on benchmark problem sets.
format Preprint
id arxiv_https___arxiv_org_abs_2109_13391
institution arXiv
publishDate 2021
record_format arxiv
spellingShingle Curvature-Aware Derivative-Free Optimization
Kim, Bumsu
Cai, HanQin
McKenzie, Daniel
Yin, Wotao
Optimization and Control
Machine Learning
49M15, 65K05, 68Q25, 90C56
The paper discusses derivative-free optimization (DFO), which involves minimizing a function without access to gradients or directional derivatives, only function evaluations. Classical DFO methods, which mimic gradient-based methods, such as Nelder-Mead and direct search have limited scalability for high-dimensional problems. Zeroth-order methods have been gaining popularity due to the demands of large-scale machine learning applications, and the paper focuses on the selection of the step size $α_k$ in these methods. The proposed approach, called Curvature-Aware Random Search (CARS), uses first- and second-order finite difference approximations to compute a candidate $α_{+}$. We prove that for strongly convex objective functions, CARS converges linearly provided that the search direction is drawn from a distribution satisfying very mild conditions. We also present a Cubic Regularized variant of CARS, named CARS-CR, which converges in a rate of $\mathcal{O}(k^{-1})$ without the assumption of strong convexity. Numerical experiments show that CARS and CARS-CR match or exceed the state-of-the-arts on benchmark problem sets.
title Curvature-Aware Derivative-Free Optimization
topic Optimization and Control
Machine Learning
49M15, 65K05, 68Q25, 90C56
url https://arxiv.org/abs/2109.13391