Fig. 3.

Illustration of the outputs of the MKL model. (A) Parietal view of the surface of the cortex of patient P1. Electrodes are displayed as circles on the cortex (not to scale) with a fill color corresponding to their contribution to the final model (d_m, in %). (B) Weights w_m for the channel circled in black in panel A. These weights display the contribution of each time point from that channel to the MKL model. As the model is not sparse in terms of the w_m, all the time points considered for modeling contribute to the final decision function.