Fast and Scalable Caputo Fractional Gradient Descent via Perturbation-Preserving Memory Compression

DOI10.48550/arXiv.2607.15505MaRDI QIDQ7284145arXivFDO

Authors Junseo Lee, Hyunju Kim, Hwanseo Lee

Publication date 16 July 2026

Published in arXiv

Tldr We propose a fast and scalable variant of Caputo fractional gradient descent that combines efficient memory compression with perturbation-preserving properties. Our new dyadic hierarchical discrete convolution (DHDC) method provides a principled way to reduce the computational cost while maintaining convergence guarantees under standard assumptions.

Full work available at URL https://doi.org/10.48550/arXiv.2607.15505

arXiv classification

cs.LGcs.NEmath.NAmath.OC

This page was built for publication: Fast and Scalable Caputo Fractional Gradient Descent via Perturbation-Preserving Memory Compression

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7284145)