Tensor-Train Decomposition

Ivan Oseledets

SIAM Journal on Scientific Computing · 2011 · 2.6K citations · 32 references

Concepts

TL;DR

The decomposition avoids the curse of dimensionality, has asymptotically the same number of parameters as the canonical form, and is stable due to low‑rank approximation of auxiliary unfolding matrices. A simple nonrecursive form of the tensor decomposition in d dimensions is presented. The authors present a fast rounding procedure and basic linear algebra operations for the decomposition. The new form enables efficient implementation of basic operations, as illustrated by computing the smallest eigenvalue of a 19‑dimensional operator.

Abstract

A simple nonrecursive form of the tensor decomposition in d dimensions is presented. It does not inherently suffer from the curse of dimensionality, it has asymptotically the same number of parameters as the canonical decomposition, but it is stable and its computation is based on low-rank approximation of auxiliary unfolding matrices. The new form gives a clear and convenient way to implement all basic operations efficiently. A fast rounding procedure is presented, as well as basic linear algebra operations. Examples showing the benefits of the decomposition are given, and the efficiency is demonstrated by the computation of the smallest eigenvalue of a 19-dimensional operator.

References

32