Lars Eldén and Maryam Dehghan

Matlab codes for Block Krylov-Schur like algorithms for

best low-rank approximation of large and sparse tensors

The codes are implementations of the algorithms developed in
A Krylov-Schur Like Method for Computing the Best Rank-(r1,r2,r3) Approximation of Large and Sparse Tensors, arXiv:2012.07595, December 2020

The codes in the packages SymKrylovSchur and NonSymKrylovSchur are written in Matlab using the Tensor toolbox (not given here) and algorithms for solving small and medium-size best low-rank tensor approximations from A Newton--Grassmann method for computing the Best Multi-Linear Rank-(r1,r2,r3) Approximation of a Tensor, SIMAX 2009. The packages grassClasses and tensorAlgs are small modifications of the two corresponding packages developed by Berkant Savas, see his web page.

The codes are distributed under a 3-Clause BSD License, see below.


February 10, 2021

Lars Eldén, Professor of Scientific Computing
Department of Mathematics
Linköping University
SE-581 83 Linköping

Email: lars 'dot' elden 'at' liu 'dot' se