search

KrystynaGrzesiak / gslope

Sparse Gaussian graphical models with Sorted L-One Penalized Estimation
https://statsimuwr.github.io/gslope
3 stars 0 forks source link

eigen() for symmetric matrices (ADMM_algorithm.R) #6

Open KrystynaGrzesiak opened 4 years ago

KrystynaGrzesiak commented 4 years ago

Is the assumption about the symmetry of C_eigen right?

mstaniak commented 4 years ago

@mmaku any thoughts?

mmaku commented 4 years ago

sorry, I forgot to answer. Yeah, I am sure that it must be symetric under some assumptions, which are always true in case of precision matrix, I remember I checked that. I'll give you theoretical proof, but give me some (little) time.

mstaniak commented 4 years ago

I see @mmaku - the problem is that @KrystynaGrzesiak found cases where it wasn't symmetric. Maybe it's a good starting point for debugging. @KrystynaGrzesiak can you provide a reproducible example of calculations failing (or being too slow?) due to non-symmetric matrix?

mmaku commented 4 years ago

We tried to set up am meeting, but unfortunately it did not work out. We will try to arrange a meeting and discuss the problem.