Structured first-layer initialization pre-training techniques to accelerate training process based on -rank

DOI10.4208/CICP.OA-2025-0185MaRDI QIDQ7267468zbMATH OpenFDO

Authors Tao Tang, Jiang Yang, Yuxiang Zhao, Quanhui Zhu

Publication date 16 June 2026

Published in Communications in Computational Physics (Search for Journal in Brave)

Full work available at URL https://doi.org/10.4208/CICP.OA-2025-0185

zbMATH Keywords

\(\varepsilon\)-rankdeep neural networkstaircase phenomenonstructured first-layer initializationtraining acceleration

Mathematics Subject Classification ID

PDEs in connection with computer science (35Q68) Applications to the sciences (65Z05) Artificial neural networks and deep learning (68T07)

This page was built for publication: Structured first-layer initialization pre-training techniques to accelerate training process based on \(\varepsilon\)-rank

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7267468)