Multivariate input vector space reconstruction and its application

This paper was concentrated on the reconstruction of multivariate input vector space. Based on evaluating the nonlinear correlation degree between observed variables and output variables, the input variables were selected if the evaluation was strong. Then, C-C method was used to reconstruct an initial input vector space. Finally, FastICA method was expanded to extract the effective independent information and reduce the dimension of initial input vector. Simulation results showed the effectiveness of the reconstructed input vector.