KernelFunctions.jl
KernelFunctions.jl copied to clipboard
Julia package for kernel functions for machine learning
In principle, it should be possible to run format checks (and build the documentation) also when a PR is opened from a fork. The security settings on Github do not...
I think our folder structure is a bit confusing. Right now we have: - kernels - basekernels - mokernels Whereas our data structure is : - Kernel - SimpleKernel -...
We could also implement this approach (Section 5) https://arxiv.org/pdf/1807.01750.pdf _Originally posted by @theogf in https://github.com/JuliaGaussianProcesses/KernelFunctions.jl/issues/245#issuecomment-766824569_
- [ ] Lengthscales & transforms (see #217) - [ ] Documentation of TensorProduct (from https://github.com/JuliaGaussianProcesses/KernelFunctions.jl/pull/213#discussion_r553540255) - [ ] Discussion around Functors.functor (from https://github.com/JuliaGaussianProcesses/KernelFunctions.jl/pull/213#discussion_r553882886)
Just realised that `Mask` was the most used name for this.
Removing unneeded definitions and using `@boundscheck` macro based on discussion in https://github.com/JuliaGaussianProcesses/AbstractGPs.jl/pull/30
#116 I have defined `kernelmatrix` for wiener kernel. This avoid the "slow method" ~~but gives `NaN`s in the Zygote gradient~~.
Once #159 is merged, we'll have a standardised set of tests in `src`. It _may_ make sense to test that these actually catch the kinds of issues that they're designed...
#159 has thrown up a couple of issues. In particular, to get the tests to pass I had to modify [this line](https://github.com/JuliaGaussianProcesses/KernelFunctions.jl/pull/159/files/e73b23d6913e43da41022159141a019150982844#diff-22b199c6f4ab07ab8a4c7af34d56a4e0R32) with an additional call to `collect`. Would be...
There are quite a lot of tests that check that `metric(kernel)` returns a particular metric. It's not clear to me how much value these tests have. It probably makes sense...