Single-Trial EEG Classification via Common Spatial Patterns with Mixed Lp- and Lq-Norms

Algorithm 1

Iterative algorithm of CSP-Lp/q for multiple spatial filters.

	Input: Training data and , norm quantities p and q, and number of spatial filters K.
	Output: Spatial filters .
(1)	Let , and set as the -dimensional identity matrix.
(2)	Compute the coefficient vector on the samples and according to the iterative procedure in Section 3.1.
(3)	Compute the spatial filter as
(4)	Deflate the samples as
	and update as
(5)	If , then let and return to Step 2. Otherwise, terminate the run and output the spatial filters .