Epilepsy is a disorder in which nerve cell activity in the brain gets disturbed, causing seizures. To accurately locate the signal source in brain cortex which caused seizure is crucial for surgery success. Stereo-EEG with multiple electrodes is a tool that used before surgery to monitor and detect electrophysiological signals from brain. In order to find out the non-linear relation between hippocampus and brain cortex with the help of stereo-EEG, kernel methods need to be involved.
In this paper, I will first explore linear adaptive filter followed by kernel adaptive filter. Kernel recursive least square algorithm with sliding-window and fixed-budget features will be implemented to make prediction of input-output pair for brain signal. The prediction will be made using FB-KRLS and the final result shows different behaviors of each channel. The prediction of channel RA and ROF indicates they have delayed signals in the same channel. While the prediction result of channel RAH and RPH indicates several electrodes will lead to the same output signal.