TT-GMRES: Solution to a linear system in the structured tensor format

S. V. Dolgov

doi:10.1515/rnam-2013-0009

TT-GMRES: Solution to a linear system in the structured tensor format

S. V. Dolgov

Russian Academy of Sciences

Research output: Contribution to journal › Article › peer-review

49 Citations (SciVal)

Abstract

An adapted tensor-structured GMRES method for the TT format is proposed and investigated. The Tensor Train (TT) approximation is a robust approach to highdimensional problems. One class of such problems involves the solution of a linear system. In this work we study the convergence of the GMRES method in the presence of tensor approximations and provide relaxation techniques to improve its performance. Several numerical examples are presented. The method is also compared with a projection TT linear solver based on the ALS and DMRG methods. On a particular SPDE (high-dimensional parametric) problem these methods manifest comparable performance, with a good preconditioner the TT-GMRES overcomes the ALS solver.

Original language	English
Pages (from-to)	149-172
Number of pages	24
Journal	Russian Journal of Numerical Analysis and Mathematical Modelling
Volume	28
Issue number	2
DOIs	https://doi.org/10.1515/rnam-2013-0009
Publication status	Published - 5 Apr 2013

Bibliographical note

Funding Information:
Russia This work was supported in part by RFBR grants 09-01-12058, 10-01-00757, 11-01-00549, RFBR/DFG grant 09-01-91332, Russian Federation Gov. contracts No. P1178, P1112 and P940, 14.740.11.0345 and Promotionstipendium of Max-Planck Gesellschaft. Part of this work was done during the stay of the author at Max-Plank Institute for Mathematics in Sciences, Leipzig, Germany.

ASJC Scopus subject areas

Numerical Analysis
Modelling and Simulation

Access to Document

10.1515/rnam-2013-0009

Cite this

@article{f4947dd3ba2f4f55adcdb75ba68ebf3e,

title = "TT-GMRES: Solution to a linear system in the structured tensor format",

abstract = "An adapted tensor-structured GMRES method for the TT format is proposed and investigated. The Tensor Train (TT) approximation is a robust approach to highdimensional problems. One class of such problems involves the solution of a linear system. In this work we study the convergence of the GMRES method in the presence of tensor approximations and provide relaxation techniques to improve its performance. Several numerical examples are presented. The method is also compared with a projection TT linear solver based on the ALS and DMRG methods. On a particular SPDE (high-dimensional parametric) problem these methods manifest comparable performance, with a good preconditioner the TT-GMRES overcomes the ALS solver.",

author = "Dolgov, {S. V.}",

note = "Funding Information: Russia This work was supported in part by RFBR grants 09-01-12058, 10-01-00757, 11-01-00549, RFBR/DFG grant 09-01-91332, Russian Federation Gov. contracts No. P1178, P1112 and P940, 14.740.11.0345 and Promotionstipendium of Max-Planck Gesellschaft. Part of this work was done during the stay of the author at Max-Plank Institute for Mathematics in Sciences, Leipzig, Germany.",

year = "2013",

month = apr,

day = "5",

doi = "10.1515/rnam-2013-0009",

language = "English",

volume = "28",

pages = "149--172",

journal = "Russian Journal of Numerical Analysis and Mathematical Modelling",

issn = "0927-6467",

publisher = "De Gruyter",

number = "2",

}

TY - JOUR

T1 - TT-GMRES

T2 - Solution to a linear system in the structured tensor format

AU - Dolgov, S. V.

N1 - Funding Information: Russia This work was supported in part by RFBR grants 09-01-12058, 10-01-00757, 11-01-00549, RFBR/DFG grant 09-01-91332, Russian Federation Gov. contracts No. P1178, P1112 and P940, 14.740.11.0345 and Promotionstipendium of Max-Planck Gesellschaft. Part of this work was done during the stay of the author at Max-Plank Institute for Mathematics in Sciences, Leipzig, Germany.

PY - 2013/4/5

Y1 - 2013/4/5

N2 - An adapted tensor-structured GMRES method for the TT format is proposed and investigated. The Tensor Train (TT) approximation is a robust approach to highdimensional problems. One class of such problems involves the solution of a linear system. In this work we study the convergence of the GMRES method in the presence of tensor approximations and provide relaxation techniques to improve its performance. Several numerical examples are presented. The method is also compared with a projection TT linear solver based on the ALS and DMRG methods. On a particular SPDE (high-dimensional parametric) problem these methods manifest comparable performance, with a good preconditioner the TT-GMRES overcomes the ALS solver.

AB - An adapted tensor-structured GMRES method for the TT format is proposed and investigated. The Tensor Train (TT) approximation is a robust approach to highdimensional problems. One class of such problems involves the solution of a linear system. In this work we study the convergence of the GMRES method in the presence of tensor approximations and provide relaxation techniques to improve its performance. Several numerical examples are presented. The method is also compared with a projection TT linear solver based on the ALS and DMRG methods. On a particular SPDE (high-dimensional parametric) problem these methods manifest comparable performance, with a good preconditioner the TT-GMRES overcomes the ALS solver.

UR - http://www.scopus.com/inward/record.url?scp=84880555884&partnerID=8YFLogxK

U2 - 10.1515/rnam-2013-0009

DO - 10.1515/rnam-2013-0009

M3 - Article

AN - SCOPUS:84880555884

SN - 0927-6467

VL - 28

SP - 149

EP - 172

JO - Russian Journal of Numerical Analysis and Mathematical Modelling

JF - Russian Journal of Numerical Analysis and Mathematical Modelling

IS - 2

ER -

TT-GMRES: Solution to a linear system in the structured tensor format

Abstract

Bibliographical note

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this