Will Tebbutt
Will Tebbutt
- [ ] Ordinal - [ ] Truncated Gaussian - [ ] Cauchy - [ ] InverseGamma - [ ] Laplace - [ ] Student-T #79 Please edit directly to...
I've always wanted to have support for ordinal regression, presumably implemented along the lines of [1]. Not particular reason -- just because it's fun. [1] - Chu, Wei and Zoubin...
This is intended as a discussion issue where we can hash out an initial design for the package. The goal is to 1. tie down the structure of the package,...
The implementation of `Matrix` and its `rrule` share quite a bit of code. We should consider how to factor some of it out into a shared implementation.
Please feel free to take charge of this PR if you want to work on this, or to make another one. I can't promise to actively work on it. I...
In an ideal world, a user would test an `rrule` by writing something like the following, and have it work all of the time: ```julia test_rrule(foo, args...) ``` By work...
Maybe it would make sense to provide a working example of testing `Zygote` using CRTU? Just working through an example, and having to dig around in Zygote internals. Should Zygote...
It's ocassionally not possible to test `@non_differentiable` using `test_rrule` because finite differencing gives the wrong / unreliable results. For example, this is the case in the `Vector{Floa64}(undef, 5)` constructor implemented...
`rand_tangent` prefers to generate natural tangents where possible. Sometimes, it's preferable to generate structural tangents. What's the feeling about the best way to accomodate this need?
It would be helpful to have functionality which tests that `@non_differentiable` stuff is properly... non-differentiable. It looks like `rand_tangent` doesn't work on it at the minute, and in any case...