bemobil_mne.preproc.compute_mi_reduction

bemobil_mne.preproc.compute_mi_reduction(raw_before, raw_after, picks='eeg')

Total pairwise MI reduction under the Gaussian approximation.

MI_total = −½ · logdet(R), where R is the channel correlation matrix. Independent channels → R = I → MI = 0. A good denoiser reduces MI by removing shared artifact variance.

Parameters:

raw_before, raw_aftermne.io.Raw

Recordings before and after denoising.

picksstr

Channel type to include (default "eeg").

Returns:

dict

mi_before, mi_after, mi_reduction (before − after), mi_reduction_pct.