Hilbert functions and tensor analysis (Q2330721)

The author gives a short introduction to the aspects of tensor analysis from the point of view of algebraic geometry and collects some results regarding tensor analysis. This is aimed for the reader that is not a specialist in algebraic geometry. We say that a tensor is identifiable if its decomposition in sum of rank \(1\) tensors is unique (up to trivialities). Let \(A\subset \mathbb{P}^n\) a non-empty set. We denote by \(l(A)\) the cardinality of \(A.\) The Kruskal rank of \(A\) is the maximaum integer \(k(A)\) such that any subset \(B\subset A\) with \(l(B)\leq k(A)\) is linearly independent. The author proves a new result: Let \(T\) be a quartic form in \(n+1\) variables, and consider a decomposition \(A\) of \(T\). Assume that \(l(A)\leq 2k(A)-1\) and that the Terracini space at \(A\) has (maximal) dimension \((2k(A)-1)(n+1)-1\). Then \(T\) has rank \(2k(A)-1\) and it is identifiable. After the main result it is proposed a algorithm to detect the identifiability of quartics of low rank. The paper ends with a discussion of future research and some open problems. For the entire collection see [Zbl 1411.81023].