An Automatic Epilepsy Detection Method Based on Improved Inductive Transfer Learning

Algorithm 2

The description of the TDLSR algorithm.

The description of TDLSR

Input:

The training samples in target domain , and their corresponding class labels , where . The maximum number of iterations is . is learned in advanced by using DLSR from source domain.

Output:

The mapping matrix and the translation vector .

Training:

Step 1: Construct the label matrix and the indicator matrix , respectively.

Step 2: Get the value of and by grid searching in the set of {10^-4, 10^-3, 10^-2, 10^-1, 10⁰, 10, 10²}, respectively.

Step 3: Initialize . Set .

Step 4: Compute by and by , respectively.

Repeat

Step 5: Update the label matrix by .

Step 6: Update and by Eqs. (17) (18).

Step 7: Update the regression error matrix by .

Step 8: Update the label shift matrix by Eq. (20).

Step 9: Let and

Until .

Step 10: Output and .