TIDE-MC: Two-Sided Interpolative Decomposition for Billion-Scale GPU Matrix Completion

DOI10.48550/arXiv.2608.00977MaRDI QIDQ7284610arXivFDO

Authors Lizheng Cheng, Fangxin Liu, Ruixuan Liu, Ming Xing Zhang, Guihai Chen, Chen Tian, Chengying Huan, Yubo Wang, Pinhuan Wang, Lizheng Chen, Jie Zhang, Qing Wang, Shaonan Ma, Mingxing Zhang, Zhibin Wang, Rong Gu

Publication date 2 August 2026

Published in arXiv

Tldr We present TIDE-MC, a GPU framework for large-scale matrix completion using Two-Sided Interpolative Decomposition (TSID), where computation scales with the anchor template rather than the complete matrix; TIDE-MC achieves up to 11,647x speedup and reduces memory usage by up to 8.5x compared to state-of-the-art GPU solvers under out-of-memory conditions.

Full work available at URL https://doi.org/10.48550/arXiv.2608.00977

arXiv classification

cs.DC

This page was built for publication: TIDE-MC: Two-Sided Interpolative Decomposition for Billion-Scale GPU Matrix Completion

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7284610)