“…In applications, the uniqueness justifies that the computed decomposition is what people wanted. Galuppi and Mella [23] showed that for a generic F ∈ S m (C n ), the Waring decomposition is unique if and only if (n, m, r) = (2, 2k − 1, 2k), (4,3,5) or (3,5,7). When F ∈ S m (C n ) is a generic tensor of a subgeneric rank r (i.e., r is smaller than the value of the Alexander-Hirschowitz formula) and m ≥ 3, Chiantini, Ottaviani and Vannieuwenhoven [8] showed that the Waring decomposition is unique, with only three exceptions: (n, m, r) = (2,6,9), (3,4,8) or (5,3,9).…”