Diagonal loading value

1

Hi dears,

I am working on regularized based MVB methods such as shrinkage based methods and conventional DL MVB method. If it is mind I have a few questions about DL MVB method:

  1. What is the DL factor in brainstorm?
  2. What is your reason or reference for selecting that DL factor value?
  3. Based on your experience, how much small one person is able to select the DL factor>
  4. Which scenario about source localization can show the advantage of shrinkage based methods in comparison to the DL based methods (with small DL factor value)?

[(In real, in my manuscript, I am getting comparable (similar) results between shrinkage method and DL MVB with DL factor equal to 0.0005landa_max in many simulated scenarios [landa_max is the maximum eigenvalue of data covariance matrix]. I don't know if this value of DL factor is appropriate so why it is not used in many publications? or if this value of DL factor is appropriate so why shrinkage based methods have been developed?), I think I could answer to these questions if I could design a scenario in which the shrinkage is superior than DL with DL factor equal to 0.0005landa_max, but currently I couldn't design such a scenario].

Any help would be appreciated.

Warm regards,
Talesh

2

@Sylvain @John_Mosher @Marc.Lalancette @MartinC @hossein27en @dimitrios @juangpc @tmedani ?

3

Dear Francois,

I'm sorry, but I cant understand your reply?

Your means is that I should wait for answers of dear people you have mentioned or I should ask them individually?

4

It means that I am not competent to address your question.
I hope one of our collaborators in this list will be able to reply.