On manifolds of tensors of fixed TT-rank

被引:139
作者
Holtz, Sebastian [1 ]
Rohwedder, Thorsten [1 ]
Schneider, Reinhold [1 ]
机构
[1] TU Berlin, D-10623 Berlin, Germany
关键词
APPROXIMATION;
D O I
10.1007/s00211-011-0419-7
中图分类号
O29 [应用数学];
学科分类号
070104 ;
摘要
Recently, the format of TT tensors (Hackbusch and Kuhn in J Fourier Anal Appl 15:706-722, 2009; Oseledets in SIAM J Sci Comput 2009, submitted; Oseledets and Tyrtyshnikov in SIAM J Sci Comput 31: 5, 2009; Oseledets and Tyrtyshnikov in Linear Algebra Appl 2009, submitted) has turned out to be a promising new format for the approximation of solutions of high dimensional problems. In this paper, we prove some new results for the TT representation of a tensor U is an element of R-n1x ... xnd and for the manifold of tensors of TT-rank (r) under bar. As a first result, we prove that the TT (or compression) ranks r(i) of a tensor U are unique and equal to the respective separation ranks of U if the components of the TT decomposition are required to fulfil a certain maximal rank condition. We then show that the set T of TT tensors of fixed rank (r) under bar locally forms an embedded manifold in R-n1x ... xnd, therefore preserving the essential theoretical properties of the Tucker format, but often showing an improved scaling behaviour. Extending a similar approach for matrices (Conte and Lubich in M2AN 44: 759, 2010), we introduce certain gauge conditions to obtain a unique representation of the tangent space TUT of T and deduce a local parametrization of the TT manifold. The parametrisation of TUT is often crucial for an algorithmic treatment of high-dimensional time-dependent PDEs and minimisation problems (Lubich in From quantum to classical molecular dynamics: reduced methods and numerical analysis, 2008). We conclude with remarks on those applications and present some numerical examples.
引用
收藏
页码:701 / 731
页数:31
相关论文
共 46 条
[1]  
[Anonymous], 1961, Adaptive Control Processes: a Guided Tour, DOI DOI 10.1515/9781400874668
[2]  
[Anonymous], 2009, Manifolds and differential geometry
[3]   The multiconfiguration time-dependent Hartree (MCTDH) method:: a highly efficient algorithm for propagating wavepackets [J].
Beck, MH ;
Jäckle, A ;
Worth, GA ;
Meyer, HD .
PHYSICS REPORTS-REVIEW SECTION OF PHYSICS LETTERS, 2000, 324 (01) :1-105
[4]   MULTIVARIATE REGRESSION AND MACHINE LEARNING WITH SUMS OF SEPARABLE FUNCTIONS [J].
Beylkin, Gregory ;
Garcke, Jochen ;
Mohlenkamp, Martin J. .
SIAM JOURNAL ON SCIENTIFIC COMPUTING, 2009, 31 (03) :1840-1857
[5]   AN ERROR ANALYSIS OF THE MULTI-CONFIGURATION TIME-DEPENDENT HARTREE METHOD OF QUANTUM DYNAMICS [J].
Conte, Dajana ;
Lubich, Christian .
ESAIM-MATHEMATICAL MODELLING AND NUMERICAL ANALYSIS-MODELISATION MATHEMATIQUE ET ANALYSE NUMERIQUE, 2010, 44 (04) :759-780
[6]   An introduction to coupled cluster theory for computational chemists [J].
Crawford, TD ;
Schaefer, HF .
REVIEWS IN COMPUTATIONAL CHEMISTRY, VOL 14, 2000, 14 :33-136
[7]   A multilinear singular value decomposition [J].
De Lathauwer, L ;
De Moor, B ;
Vandewalle, J .
SIAM JOURNAL ON MATRIX ANALYSIS AND APPLICATIONS, 2000, 21 (04) :1253-1278
[8]   On the best rank-1 and rank-(R1,R2,...,RN) approximation of higher-order tensors [J].
De Lathauwer, L ;
De Moor, B ;
Vandewalle, J .
SIAM JOURNAL ON MATRIX ANALYSIS AND APPLICATIONS, 2000, 21 (04) :1324-1342
[9]   TENSOR RANK AND THE ILL-POSEDNESS OF THE BEST LOW-RANK APPROXIMATION PROBLEM [J].
de Silva, Vin ;
Lim, Lek-Heng .
SIAM JOURNAL ON MATRIX ANALYSIS AND APPLICATIONS, 2008, 30 (03) :1084-1127
[10]   A NEWTON-GRASSMANN METHOD FOR COMPUTING THE BEST MULTILINEAR RANK-(r1, r2, r3) APPROXIMATION OF A TENSOR [J].
Elden, Lars ;
Savas, Berkant .
SIAM JOURNAL ON MATRIX ANALYSIS AND APPLICATIONS, 2009, 31 (02) :248-271