Input: Data set ; the reduced dimension d, parameter p, . Output: Coefficient matrix . 1: Construct the kernel matrix ; 2: Decompose by using eigenvectors decomposition; 3: Initialize by using KDA [42] ; 4: Repeat step 5-step 9; 5: Update base on (13) ; 6: Update ; 7: Update by using the eigenvalue decomposition of (18); 8: Update by ; 9: ; 10: Until convergence; 11: Return coefficient matrix . |