Will Tebbutt

I wonder whether something like ```julia approx = VFE(f(z), f(x), y) elbo(approx) # returns a scalar approx_posterior = posterior(approx) # return an ApproxPosteriorGP ``` and ```julia approx = VFE(f(z), f(x),...

Yeah -- I at least think we ought to be able to do this in the deterministic-objective case (no minibatching, quadrature / exact reconstruction-term computation) at the very least. I've...

> For example, it seems rather inefficient that we have to compute posterior(fx, y_train) all over whenever we want to plot... isn't there some way to get it once together...

Ooo not come across them before. I'll add reading them to my TODO list. AbstractGPs is more of an interface package. For example, I'm going to refactor [Stheno.jl](https://github.com/willtebbutt/Stheno.jl/) and [TemporalGPs.jl](https://github.com/willtebbutt/TemporalGPs.jl)...

I agree that it would be great to have a version of your celerite work that implements the `AbstractGP` API, which should be very doable from what I've understood from...

One large area that we've really not covered so far is benchmarking. It would be very helpful to have a systematic suite of both "unit benchmarks" and "integration benchmarks" that...

> Is your goal to use this package as a framework for writing packages like Stheno and AugmentedGPs? And if so, are you planning to refactor those packages accordingly? This...

This sounds like a good idea. I guess this would be straightforward, since we've been writing our examples using Literate anyway, right?

Another thing that we might consider doing is provide examples that use [ParameterHandling.jl](https://github.com/invenia/ParameterHandling.jl/) and `Functors.jl` interface.

I've closed #30 because it went stale. @thomasgudjonwright volunteered to take a look at the general problem of docs for MOGP some time next month (hopefully). @sharanry if you get...