General tensor decomposition, moment matrices and applications (Q1940930)

This article studies the tensor decomposition problem: Given a partially symmetric tensor \(T(x)\in S^{\delta_1}(E_1)\otimes\cdots\otimes S^{\delta_k})(E_k)\), find a decomposition of \(T(x)\) as a sum of \(r\) tensors of rank \(1\). Here \(r\) is the smallest possible number and is called the rank \(T\). Ultimately, the problem is transformed into the truncated moment problem: Given a partially symmetric tensor \(T(x)\), associate it with \(T^*\in R^*_{\delta_1\cdots\delta_k}\), find the smallest \(r\) such that there exists \(\Lambda\in R^*\) which extend \(T^*\) with the moment matrix \(H_\Lambda\) of rank \(r\). The article then describes an algorithm, which is an extension of the authors' old results, to calculate the rank and the decomposition of \(T\). Examples are given in the last section.