On the convergence of Jacobi-type algorithms for Independent Component Analysis

Jianze Li, Konstantin Usevich, Pierre Comon

Abstract: Jacobi-type algorithms for simultaneous approximate diagonalization of real (or complex) symmetric tensors have been widely used in independent component analysis (ICA) because of their good performance. One natural way of choosing the index pairs in Jacobitype algorithms is the classical cyclic ordering, while the other way is based on the Riemannian gradient in each iteration. In this paper, we mainly review in an accessible manner our recent results in a series of papers about weak and global convergence of these Jacobitype algorithms. These results are mainly based on the Lojasiewicz gradient inequality.