Model reduction through multilinear singular value decompositions

Model reduction through multilinear singular value decompositions

Weiland, S.
Van Belzen, F.

2006-09-07

This paper considers the problem of optimal rank approximation of multilinear functions. A new notion of singular values is introduced for an arbitrary tensor and it is shown how optimal lower rank approximations of a tensor can be inferred from these singular values, without resorting to orthogonal tensor decompositions. The results in this paper are primarily motvated by the problem to find optimal projection spaces for model reduction purposes. It is shown that the approximation results outperform earlier singular value based techniques of lower rank approximations of tensors.

singular values
multilinear algebra
POD model reduction
tensor calculus
optimal rank approximation

http://resolver.tudelft.nl/uuid:d628abb7-16fc-4778-913c-f9564f6475e4

Conference proceedings

conference paper

(c) 2006 Weiland, S.; Van Belzen, F.