{"id":"arxiv/2610.16649","url":"https://arxiv.org/abs/2610.16649","tags":["math.NA"],"score":0,"title":"Globally Convergent Jacobi-Type Algorithms for Robust Approximate Diagonalization of Higher-Order Cumulant Tensors","domain":"math.NA","source":"arxiv","authors":["Jianze Li","Zhenyu Cai","Wentao Ding","Konstantin Usevich"],"reviews":0,"subject":"math.NA","summary":"The sums of the absolute values of higher-order cumulants are classical contrast functions for independent component analysis (ICA) and lead naturally to robust approximate diagonalization of higher-order cumulant tensors using $\\ell_1$-based diagonalization over the orthogonal or unitary group. We propose Jacobi-SubG, a Jacobi-type framework that selects index pairs using Riemannian subdifferential information and performs plane rotations with a refinement step that preserves nonzero diagonal contractions. For the general tensor model, we prove global convergence of the entire iterate…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-12","reactions":{}}