Pseudo-marginal MCMC for non-conjugate GP inference

[Red-Portal]

Hi,

I have personally experienced that pseudo-marginal MCMC (PM-MCMC) methods are quite efficient for low-dimensional non-conjugate GPs. (See this paper.) Since I have some experience implementing them, I would be interested to contribute an implementation of PM slice sampling somewhere, but it's kinda tricky where this will have to go. My first guess is this repo since it contains Laplace approximation, which is a prerequisite for PM-MCMC on GPs; what do you guys think?