Low-Rank Triple Decomposition of Streaming Tensors and Its Application to Video Completion
Triple tensor decomposition (TriTD) has recently emerged as a good alternative for the two classical CP/PARAFAC and Tucker tensor decompositions. However, current state-of-theart TriTD methods are primarily designed for batch processing, requiring the storage and processing the entire tensor data at once. In this paper, we propose, for the first time, an effective and efficient incremental method for online tensor decomposition in the TriTD format. We validate the performance of our method with the task of online video completion. Experimental results demonstrate that our method outperforms state-of-the-art streaming tensor decomposition methods.
Authors
- Olivier Buttelli (ORCID: https://orcid.org/0000-0001-7290-6344)
- Karim Abed‐Meraim (ORCID: https://orcid.org/0000-0003-2652-1923)
- Thanh Trung Le
- Ha Vu Le (ORCID: https://orcid.org/0000-0003-3441-1409)
- Philippe Ravier (ORCID: https://orcid.org/0000-0002-0925-6905)
- Hồng Quân Nguyễn (ORCID: https://orcid.org/0000-0001-7685-8191)
- Ha Manh Luu (ORCID: https://orcid.org/0000-0002-5678-6212)
- Thi Thuy Quynh Tran (ORCID: https://orcid.org/0009-0009-2996-1853)
- Vy-Thuy-Lynh Hoang
Institutions
- Université d'Orléans (FR)
- Vietnam National University, Hanoi (VN)
- China Geological Survey (CN)
- Laboratoire Pluridisciplinaire de Recherche en Ingénierie des Systèmes, Mécanique et Energétique (FR)
- Bureau de Recherches Géologiques et Minières (FR)
Publication Details
- Published
- 2025-10-16
- DOI
- https://doi.org/10.1109/iscit67082.2025.11231620
- Primary Topic
- Tensor decomposition and applications
- Type
- article
- Field-Weighted Citation Impact
- 0.00