Seth Axen
Seth Axen
With package extensions, perhaps it would be possible to add a SpecialFunctionsChainRulesCoreHypergeometricFunctionsExt that defines the rules with respect to the first argument. Then either a fallback rule could be added...
> Is this really possible? HypergeometricFunctions depends on SpecialFunctions, so I wonder if it can be a weak dependency at all (I am sure it can't be a proper dependency...
Thanks for the pointer, @lrnv . Agreed it is likely faster to differentiate through a hypergeometric function if supported than to implement the derivative separately. @devmotion any objections to the...
The issue here seems to be that if `ess`, `rhat`, and `mcse` all fail, then `nt_additional` is empty, and this line tries to create an empty `ChainDataFrame`, which is not...
MCMCDiagnosticTools v0.3.10 contains a fix for this: https://github.com/TuringLang/MCMCDiagnosticTools.jl/pull/119
I would suggest waiting on this until #430 is completed, as this will make some breaking changes to `hpd`, deprecate it, and replace it with `PosteriorStats.hdi`
Yeah it's pending a breaking changes to PosteriorStats I need to finish up, to avoid another breaking release here.
@devmotion @torfjelde This is ready for final review.
Currently PosteriorStats has no `try/catch` mechanism for if a given statistic fails. That causes chains with less than 10 draws to error upon computation of summary stats (or display). For...
This is ready for another review. Also, it would be nice to get input on https://github.com/arviz-devs/PosteriorStats.jl/issues/25, since that's relevant for this PR.