Seth Axen

@mattjj no, it's not a priority. I do need repeated eigenvalues, but I'm only test driving jax for the moment while doing my main work with a different system. Feel...

FWIW, I originally encountered this while playing around with the matrix exponential of hermitian matrices, which is a power series function. For such functions that internally use the eigendecomposition, we...

> It looks like we could use the forward derivative for a JVP rule in JAX, which would suffice for auto-diff as long as we know how to implement and...

I spent some time going through the paper today, and as far as I can tell, this approach only handles exactly degenerate matrices, not almost-degenerate matrices, and in the case...

@mlindauer we spoke about this briefly at the AutoML fall school, and you said it might be straightforward to add. If so, I could try to contribute this feature, but...

Thanks @mlindauer for the pointers! These were very helpful. I was able to put together an example that seems to work well on some toy target functions with unknown constraints:...

> We would appreciate it if you could provide a PR for us (ideally for [development 2.0 branch](https://github.com/automl/SMAC3/tree/development)) (you can also come to us if you need any support with...

Hi @MasonProtter , has there been any work toward fixing this in https://github.com/JuliaFolds2/Transducers.jl?

The following packages already call `varcorrection` directly: - GeoStatsBase.jl - Manifolds.jl - Microeconometrics.jl GeoStatsBase also overloads `varcorrection` for its own `GeoWeights

Yes, I can look into this. @baggepinnen I'm not familiar with ForwardDiffChainRules. Can you provide an MWE that exhibits the observed failure using just ChainRules? Using our own testing machinery,...