ApproximateGPs.jl

ApproximateGPs.jl copied to clipboard

Add Random Fourier Features

3

Might it be possible to add the random Fourier features approximation? Here's a draft pull request.

samanklesaria

build on `approx_log_evidence` introduced in AbstractGPs 0.5.17

4

Deprecates the `approx_lml` defined in this package in favour of `AbstractGPs.approx_log_evidence` (see https://github.com/JuliaGaussianProcesses/AbstractGPs.jl/pull/361).

st--

[DO NOT MERGE] Use CompatHelper from KernelFunctions

2

This uses the much more detailed and advanced CompatHelper script from KernelFunctions.jl

theogf

`_prior_kl` not differentiable for `Centered` with `Zygote`

14

Here is simplified view of the problem from @simsurace: ```julia using Distributions using Zygote function DKL(par1, par2) K1 = [par1[1] par1[2]; par1[2] par1[1]] K2 = [par2[1] par2[2]; par2[2] par2[1]] return...

theogf

GPR/VFE equivalence checks fail on nightly

1

See https://github.com/JuliaGaussianProcesses/ApproximateGPs.jl/runs/7748452163?check_suite_focus=true ``` Test Summary: | Pass Fail Error Broken Total Time ApproximateGPs | 181 1 1 2 185 2m25.0s sparse_variational | 112 1 1 114 1m22.1s AbstractGPs interface |...

simsurace

Hi, I have personally experienced that pseudo-marginal MCMC (PM-MCMC) methods are quite efficient for low-dimensional non-conjugate GPs. (See [this paper](https://ieeexplore.ieee.org/document/6786502).) Since I have some experience implementing them, I would be...

Red-Portal

Factor out homoscadasticity check in elbo

1

Relaxes the type signature of `elbo` to make it possible to extend

rossviljoen

Include e.g. why we have to pass a jitter to `LatentGP` (see [comment](https://github.com/JuliaGaussianProcesses/ApproximateGPs.jl/pull/59/files?authenticity_token=eenE3GK8QMpUTnVNnhDwFD7lQnVjmDHueNjES3WJdJUa3E0XLSz7mfYQqbZsDB2dg0qylnT8VXX069chAoaviQ%3D%3D&file-filters%5B%5D=.jl&file-filters%5B%5D=.toml#r715694354))

st--

Support for non-zero prior mean

1

We currently don't have any tests that a non-zero prior mean actually works. - [X] SparseVariationalApproximation: supported (with a bug fixed in #86) - [ ] LaplaceApproximation: might or might...

st--

WIP: bugfix AD

1

st--