TT Decomposition
TT: Tensor Train
Overview
Tensor-train decomposition (TT) represents a high-dimensional tensor as a sequence of smaller cores joined by rank-bounded links, reducing storage when the data has exploitable low-rank structure. Methods such as TT-SVD and TT-cross construct these representations, while ALS and DMRG optimize their cores or adapt their ranks. Researchers use TT models for high-dimensional PDEs and dynamics, probabilistic inference, signal and image processing, and compression of neural networks. A central challenge is choosing ranks and mode orderings that preserve accuracy: compression and speed depend on the data’s structure, and forcing low rank along poorly compressible modes can degrade results.
Latest papers
22 papers1 paper in the last four weeks, down 67% on the four weeks before. 0.0% of all new papers.