Spherical-Design Oracles for Zeroth-Order Optimisation: Deterministic, Variance-Free Gradient Estimation

DOI10.48550/ARXIV.2610.24373MaRDI QIDQ7400172arXivFDO

Authors Iman Shames, Amir Ali Farzin, Yuen-Man Pun, Philipp Braun

Publication date 17 October 2026

Published in arXiv

Tldr We develop a zeroth-order optimization method that estimates the gradient with minimal bias and variance by using central differences on directions given by a fixed spherical t-design. This oracle is deterministic, has no variance, and we demonstrate its superior performance in numerical experiments.

Full work available at URL https://doi.org/10.48550/ARXIV.2610.24373

arXiv classification

cs.GTmath.NAmath.OC

This page was built for publication: Spherical-Design Oracles for Zeroth-Order Optimisation: Deterministic, Variance-Free Gradient Estimation

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7400172)