Error-controlling algorithm for simultaneous block-diagonalization and its application to independent component analysis

摘要

The finest simultaneous block-diagonalization for a given set of square matrices has been studied independently in the area of independent component analysis (ICA) and semidefinite programming. A new algorithm for this problem, which finds the finest decomposition with a capability of coping with numerical errors, has recently been proposed by the present authors. In this paper we indicate the use of the algorithm for ICA by describing its main features and comparing the method with the other existing methods.