Online TT-ALS for Streaming Tensor Decomposition with Incremental Orthogonalization

DOI10.48550/arXiv.2606.31061MaRDI QIDQ7270415zbMATH OpenarXivFDO

Authors Yuto Miyatake, Daisuke Furihata, Hiroki Takeda

Publication date 30 June 2026

Published in arXiv

Tldr We develop an Online TT-ALS algorithm that leverages incremental orthogonalization to efficiently compute tensor train decompositions from streaming data with guaranteed convergence and low memory footprint, outperforming existing online methods in terms of both mathematical accuracy and human perception-based video quality metrics.

Full work available at URL https://doi.org/10.48550/arXiv.2606.31061

Mathematics Subject Classification ID

Multilinear algebra, tensor calculus (15A69) Numerical methods for low-rank matrix approximation; matrix compression (65F55)

arXiv classification

cs.CVcs.LGmath.NA

Cites work

• Efficient Tensor Completion for Color Image and Video Recovery: Low-Rank Tensor Train
• Image Quality Assessment: From Error Visibility to Structural Similarity
• Subspace Learning and Imputation for Streaming Big Data Matrices and Tensors
• Tensor Decompositions and Applications
• Tensor-train decomposition
• The alternating linear scheme for tensor optimization in the tensor train format

This page was built for publication: Online TT-ALS for Streaming Tensor Decomposition with Incremental Orthogonalization

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7270415)