david-vicente

[Request] Huber function, concomitant scale estimation

9

The Matlab version of CVX [1,2] offers the options of calling the Huber function with either `huber(x,M)` or `huber(x,M,t)`. The latter computes `t+t*huber(x/t,M)`, which is useful for concomitant scale estimation...

[Request] Better navigation in a julia buffer.

10

Right now in a `julia-mode` buffer neither `forward-paragraph` nor `forward-sexp` (and their `backward` equivalents) understand julia syntax to move between function definitions and control-flow code blocks. For example, the default...

`PeriodicKernel` does not work with AD (see issue for work-around)

10

Hi. I noticed that the PeriodicKernel is failling AD tests and I would like to implement the Mauna Loa GPR example. I tried looking for discussion regarding this issue in...

Add `with_period` convenience function and a example to `PeriodicTransform` docstring

9

This PR follows the suggestion [in this comment](https://github.com/JuliaGaussianProcesses/KernelFunctions.jl/issues/389#issuecomment-961956874 ) As it is `with_period` only works with scalars, which contrasts with `with_lengthscale` that uses `ARDTransform` to deal with vector inputs. Any...

Now that testing methods for MOKernels ara available maybe we can improve the code here: https://github.com/JuliaGaussianProcesses/KernelFunctions.jl/blob/4569859cef362ef3c1fe3848290bfb090a7ee0f1/test/mokernels/slfm.jl#L33

Maybe this type of Kernels should all implement such a method. Following this comment https://github.com/JuliaGaussianProcesses/KernelFunctions.jl/pull/263#discussion_r641687524 .

[Question] Train a multi-input multi-output model with corregionalization.

3

I managed to train a multi-input single-output with no issues. Consider the fake data: ``` Julia x1 = -5:0.8:5 x2 = -5:0.8:5 xx1 = vec([i for i in x1, j...

Wrap some $ characters with org-mode markup for verbatim. Github seems to try and render Latex, and was wrongly parsing some of the text in this file as Latex.

- Previously we couldn't use NonNegQuadReg because it was not declared a subtype of Regularizer, as GLRM methods' signatures require it to be. - The `evaluate` method for NonNegQuadReg used...