Open jamesrobertlloyd opened 11 years ago
Hessian delta of 1e-6 will not always be appropriate e.g. location parameters for unscaled data
The default prior in the laplace approx needs to be revisited
NaNs for scores mess up sorting routines - affects the printing of results I think
Hessian delta of 1e-6 will not always be appropriate e.g. location parameters for unscaled data