Gradient descent fails to learn high-frequency functions and modular arithmetic

DOI10.1007/S10994-025-06747-8MaRDI QIDQ6984724zbMATH OpenFDO

Authors Rustem Takhanov, Maxat Tezekbayev, Artur Pak, Arman Bolatov, Zhenisbek Assylbekov

Publication date 14 May 2025

Published in Machine Learning (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1007/S10994-025-06747-8

zbMATH Keywords

barren plateaugradient-based optimizationhardness of learninghigh-frequency periodic functionsmodular multiplicationstatistical query

Mathematics Subject Classification ID

Miscellaneous applications of number theory (11Z05) Computational difficulty of problems (lower bounds, completeness, difficulty of approximation, etc.) (68Q17) Learning and adaptive systems in artificial intelligence (68T05) Applications of mathematical programming (90C90)

Cites work

• A General Moment Problem
• Almost orthogonal series
• Cryptographic hardness for learning intersections of halfspaces
• Distribution-specific hardness of learning neural networks
• Efficient noise-tolerant learning from statistical queries
• scientific article; zbMATH DE number 2080649 (Why is no real title available?)
• Many \(T\) copies in \(H\)-free graphs
• Optimal Young's inequality and its converse: A simple proof
• Statistical query algorithms for mean vector estimation and stochastic convex optimization
• Unconditional lower bounds for learning intersections of halfspaces
• Weakly learning DNF and characterizing statistical query learning using Fourier analysis

Cited in

(1)

• The informativeness of the gradient revisited

This page was built for publication: Gradient descent fails to learn high-frequency functions and modular arithmetic

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6984724)