Independent component ordering in ICA time series analysis

Independent component analysis (ICA) has provided a new tool to analyze tim e series, which also gives rise to a question - how to order independent co mponents? In the literature, some methods (Back and Trappenberg, Proceeding s of International Joint Conference on Neural Networks, Vol. 2, 1999 pp. 98 9-992; Hyvarinen, Neural Computing Surveys 2 (1999) 94; Back and Weigend, I nt. J, Neural Systems 8(4) (1997) 473) have been suggested to decide the or der based on each individual component without considering their interactio ns on the observed series. In this paper, we propose an alternative approac h to order the components according to their joint contributions in data re construction. which naturally leads the component ordering to a typical com binatorial optimization problem, whereby the underlying optimum ordering ca n be found in an exhaustive way. To save computing costs., we also present a fast approximate search algorithm. The accompanying experiments have show n the outperformance of this new approach in comparison with an existing me thod. (C) 2001 Elsevier Science B.V. All rights reserved.