Multifeature Metric Learning Based on Enhanced Equidistance Embedding for Electroencephalogram Recognition of Epilepsy

Algorithm 1

MMLE³.

Input: Multi-feature matrix , its cannot-link and must-link sets;

Output: The metric matrix Q and .

Initialization:, Q = I,

Repeat

t = t+1

Step 1: fixed Q(t),compute using Eq.(13);

Step 2: fixed , compute Q(t) using Eqs.(9)-(11);

Step 3: compute the value of objective function J(t);

Until J(t) is convergence or

Step 4: obtain the optimal Q and .