Firstly, thank you for the great library the is MNE.
My question is regarding the infomax implementation.
At the original paper of Lee et al. (1999) they use a guaranteed stable algorithm for switching the k parameters in order to reestimate the number of subgaussian sources (equation 2.28).
In MNE’s implementation it uses an empirical calculation of the recovered sources kurtosis (which are considered positive up to a
signsbias value) with momentum. As I am using MNE for my Master’s degree research I would like to kindly ask for the reasoning of this method, why the proposed method in the paper was not used (as it at least seems to be easy to implement) and some reference, if possible.
- MNE version: 1.0