Regularized Embedded Multiple Kernel Dimensionality Reduction for Mine Signal Processing

Algorithm 1

The proposed MKL-EGE algorithm.

Input: The matrix of data points , the number of classes , step length , maximum number of iterations ,

   parameters , and , an error constant .

Output:

() Initialize , construct the weighted matrix , calculate .

() Repeat

   () Calculate and .

   () Find as the first c largest eigenvalues of and   as the first c smallest eigenvalues of .

   () Let , and .

   () while   do

     () Compute as the sum of the first c largest eigenvalues of .

     () If    then    else  .

     () .

   () Obtain , where are the c eigenvectors corresponding to the c largest eigenvalues of .

   () Set  .

   () Update .

    until   reached

() Output , calculate the embedding result