st--

> Positive diagonal means it's a valid cholesky factor, so can just be plugged into stuff that expects that kind of constraint, such as a `Cholesky` (i.e. constructing it directly,...

Hi, I have to admit I might not have thought through well enough how to use it in the scalar case (1-dimensional input) ... The code generally assumes that it's...

The challenge is how to distinguish between "vector = a single data point with multiple covariates" and "vector = multiple data points with a single covariate each"... (e.g. KernelFunctions.jl exports...

like a bunch of other people I added an update before checking whether someone else had already done so :sweat_smile: so this one is overlapping with #712, #721, and #708.

There might be scope for encapsulating something like what Will posted within an interface _similar_ to `rand(f)`, but as for approximate inference in non-conjugate GP models, there's different ways one...

(The "call" API appears like it ought to be stateful, and one downside of the Cholesky approach is that this is not the case - each subsequent call becomes more...

You should also call the kernel object directly, not its `.K` method. Correct would be: `kernel(data)`