Tensor-stabilizer characterization for forests

Let GG be a forest on vertex set V=1,,dV=\\{1,\ldots,d\\}, with each connected component a tree having at least three nodes, and let AA be its adjacency matrix. Let VG\mathcal{V}_G be the variety of symmetric tensors whose entries vanish whenever the subgraph induced by their indices is disconnected. Denote by O(d)O(d) the orthogonal group, by SP(d)SP(d) the signed permutation matrices, and by P(d)\mathcal{P}(d) the permutation matrices. For QO(d)Q\in O(d), let GT(VG)\mathcal{G}_T(\mathcal{V}_G) denote the tensor symmetry group of VG\mathcal{V}_G. Forest stabilizer conjecture. Let QO(d)Q\in O(d). Then QGT(VG)Q\in\mathcal{G}_T(\mathcal{V}_G) if and only if QSP(d)Q\in SP(d) and PTAP=AP^TAP=A, where PP(d)P\in\mathcal{P}(d) is a permutation matrix such that Pij=QijP_{ij}=|Q_{ij}|. This would characterize all orthogonal transformations preserving the tensor zero structure induced by such a forest.

Sources & referencesView supporting material

Primary source

Marina Garrote-López and Monroe Stephenson, “Cumulant Tensors in Partitioned Independent Component Analysis”, arXiv:2402.10089 (2024).

Progress summary

Never refreshed

Nothing recorded yet. Refresh searches the literature and the public web for attempts on this problem, and writes the first summary here.

Solutions 0

No solutions have been posted yet.