High-Probability Guarantees for Random Zeroth-Order Gradient Descent on Smooth Functions

Haishan Ye · 2026

Read High-Probability Guarantees for Random Zeroth-Order Gradient Descent on Smooth Functions by Haishan Ye (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-26547.txt.

Description

Haishan Ye studies a black-box search method that tunes settings using only output values, never gradients. Past high-chance claims came from a rough conversion that forced a smoothing width to shrink as confidence grew. Ye instead tracks the search path step by step and proves sharp odds of landing near the best point. The bounds cover strongly curved, plain curved, and bumpy smooth goals, with query counts spelled out for each case.

Catalog record

CREATORHaishan Ye
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE91555
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.26547

Browse the full library · About the Tower of Babel