KERNEL MAIN COMPONENT ANALYSIS METHOD, KERNEL MAIN COMPONENT ANALYZER, KERNEL MAIN COMPONENT ANALYSIS PROGRAM

摘要

PROBLEM TO BE SOLVED: To provide a kernel main component analysis method capable of reducing a calculation amount, and capable of reducing lowering of calculation precision getting low in accompaniment to the reduction of the calculation amount, compared with conventional one.;SOLUTION: This main component analysis method includes a selection step for selecting the m-number of sample vectors {y₁, y₂ to y_m} from n-number of sample vectors {x₁, x₂ to x_n}, a design step for calculating the r-number of eigenvectors {z₁, z₂ to z_r}, by solving a generalized eigenvalue problem K_xy^TK_xyz=K_yz, and an evaluation step for calculating a feature vector y=[z₁, z₂ to z_r]^Tx _{of an evaluation-objective vector x.;COPYRIGHT: (C)2011,JPO&INPIT}