JuliaGaussianProcesses/ApproximateGPs.jl

Laplace objective.f

Opened this issue · 1 comments

Having slept on this, I'm still not completely comfortable with suggesting that the official way to get f is via the objective because there seem to be details that change depending upon where it's defined.

Maybe it would make sense to add a struct which explicitly contains f, and whose evaluation is defined to compute the objective? i.e. formalise the API.

Yeah I was wondering something similar. would you be up for proposing a way of doing that in a separate PR?

Originally posted by @st-- in #59 (comment)

st-- commented

I just realised I could've resolved this with #109 in one go 🤦