Polyak Stepsize: Estimating Optimal Functional Values Without Parameters or Prior Knowledge

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Abdukhakimov, Farshed, Pham, Cuong Anh, Horváth, Samuel, Takáč, Martin, Hanzely, Slavomır
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866908500154646528
author Abdukhakimov, Farshed
Pham, Cuong Anh
Horváth, Samuel
Takáč, Martin
Hanzely, Slavomır
author_facet Abdukhakimov, Farshed
Pham, Cuong Anh
Horváth, Samuel
Takáč, Martin
Hanzely, Slavomır
contents The Polyak stepsize for Gradient Descent is known for its fast convergence but requires prior knowledge of the optimal functional value, which is often unavailable in practice. In this paper, we propose a parameter-free approach that estimates this unknown value during the algorithm's execution, enabling a parameter-free stepsize schedule. Our method maintains two sequences of iterates: one with a higher functional value is updated using the Polyak stepsize, and the other one with a lower functional value is used as an estimate of the optimal functional value. We provide a theoretical analysis of the approach and validate its performance through numerical experiments. The results demonstrate that our method achieves competitive performance without relying on prior function-dependent information.
format Preprint
id arxiv_https___arxiv_org_abs_2508_17288
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Polyak Stepsize: Estimating Optimal Functional Values Without Parameters or Prior Knowledge
Abdukhakimov, Farshed
Pham, Cuong Anh
Horváth, Samuel
Takáč, Martin
Hanzely, Slavomır
Optimization and Control
The Polyak stepsize for Gradient Descent is known for its fast convergence but requires prior knowledge of the optimal functional value, which is often unavailable in practice. In this paper, we propose a parameter-free approach that estimates this unknown value during the algorithm's execution, enabling a parameter-free stepsize schedule. Our method maintains two sequences of iterates: one with a higher functional value is updated using the Polyak stepsize, and the other one with a lower functional value is used as an estimate of the optimal functional value. We provide a theoretical analysis of the approach and validate its performance through numerical experiments. The results demonstrate that our method achieves competitive performance without relying on prior function-dependent information.
title Polyak Stepsize: Estimating Optimal Functional Values Without Parameters or Prior Knowledge
topic Optimization and Control
url https://arxiv.org/abs/2508.17288