GP regularization effect

Yes it does look like you’re getting those errors because one or more of the etas is going to zero. Maybe you could try using Gammas or Lognormals for the eta parameters? Or other boundary avoiding priors?

1 Like