AbstractGPs.jl

AbstractGPs.jl copied to clipboard

Stuff that needs doing

7

@4aHxKzD was asking about ways to contribute to this package in a different issue, which made me realise that we don't currently have a TODO list for this package, or...

willtebbutt

Celerite methods?

8

Anyone here familiar with the methods used [here](https://arxiv.org/abs/1703.09710)? It gives a linear-time exact solution for special case of one-dimensional data with covariance given by mixtures of exponentials. The main implementation...

cscherrer

rrule for MeanFunction types

5

Without `rrule` AD will not pass for `MeanFunction` types. #16 resolves #14 for `ZeroMean` but it should be done for all types.

theogf

Kernel for multidimensional output

3

I wanted to implement the Gaussian Process Implicit Surface (GPIS) algorithm in Julia using your package but was unable to figure out I could code a kernel for multi-dimensional outputs....

jakiroshah

Create Downgrade-CI and bump lower compat bounds

10

Following the general suggestion [here](https://discourse.julialang.org/t/psa-add-downgrade-ci-to-better-check-version-compatability/110063). I'm trying the workflow file from [here](https://gist.github.com/aplavin/ec1024d00db6ebe8a7fa0a38aa8cca6c) to speed up finding bad lower bounds. The PR raises lower bounds such that if any single one...

simsurace

Zygote errors with parameterized mean functions and multidimensional input

10

When trying to differentiate `logpdf` or other scalar functions with a parameterized mean function and multidimensional input, there are errors: ```julia using AbstractGPs using Zygote pars = [1., 0.] function...

simsurace

Hyperparameter optimization & maintanance

7

Hi! Thank you for this wonderful package. I am implementing a Trust Region Bayesian Optimization (TuRBO) algorithm during Google Summer of Code, and would like to depend on AbstractGPs. A...

samuelbelko

Noise parameter `Sigma` can't be recovered from `PosteriorGPs`

5

I noticed that once you have a `PosteriorGP`, it is no longer possible to recover the noise level that was given in the construction of the `FiniteGP`. My quick fix...

Crown421

An abstraction for the realization of a GP

11

So in the documentation there is the sentence: > An AbstractGP, f, should be thought of as a distribution over functions. This means that the output of rand(f) would be...

FelixBenning

In order to design a front-end that is at least as flexible as GaussianProcesses.jl, and to check that we are at feature parity currently (albeit with more boilerplate), we are...

willtebbutt