Thank you; that was helpful. One quick follow-up: In reference [5],
weight_norm is set to None, but before the beamformer filters are
calculated, the leadfield matrix column vectors are normalised to unity. If
I understand correctly, this is equivalent to array-gain weight
normalisation (which is as far as I know not directly implemented in
MNE-Python beamformer approaches). Is that correct?
maybe I can chime in here.
Yes, normalizing the lead field is also known as the array-gain beamformer
(you basically normalize the lead field instead of the beamformer weights
as another way to control the center-of-head bias). This is implemented in
the DICS as the parameter "normalize_fwd".
Since you asked about "beamformer approaches", note that in the LCMV code
it is the "depth" parameter that controls this - and this will be unified
among beamformers with the effort Marijn described in his email earlier.