David Widmann

> > No, it's not supported by `ess_rhat`. All functions either work with vectors or arrays of shape (draws, chains, params), but not both. > > I believe `ess_rhat` currently...

In my opinion, we should not "unify" functions that operate on single chains by moving them to an interface that works on 3d arrays of multiple chains with multiple parameters....

It seems a major difference is that glmnet 4.0 can fit any GLM family, see, e.g. https://statisticaloddsandends.wordpress.com/2020/05/14/glmnet-v4-0-generalizing-the-family-parameter/ and https://cran.r-project.org/web/packages/glmnet/vignettes/glmnetFamily.pdf.

From the blog post I got the impression that this generalization was only possible by generalizing the Fortran code as well: > Before v4.0, glmnet() could only optimize the penalized...

I've encountered this smoothed version in OT (e.g., https://papers.nips.cc/paper/2016/file/2a27b8144ac02f67687f76782a3b5d8f-Paper.pdf) but just used `logsumexp` with an iterator to implement it (https://github.com/devmotion/StochasticOptimalTransport.jl/blob/7960dc8668c27236da3b0429a86fb2ad57f9c382/src/utils.jl#L111-L113, seems I should switch to LogExpFunctions now it's reexported from...

Would it be just ```julia logsumexp(f, x) = logsumexp(f(xi) for xi in x) ``` ? Or maybe ```julia logsumexp(f, x; dims=:) = logsumexp(Broadcast.instantiate(Broadcast.broadcasted(f, x)); dims=dims) ``` + changing `::AbstractArray` to...

> @devmotion I guess we also need some ChainRules methods... We could, but it's not necessary to do in this PR IMO. Apart from the element type (see discussion above),...

> What's left to do? You have to add the function to docs/src/index.md.

> although for large enough arrays it gives Inf I guess it should depend on the range of values in the array and could possibly happen even with very small...

> In GLM.jl, the gamma residual deviance is computed as A bit unrelated, but using LogExpFunctions this could be implemented as ```julia devresid(::Gamma, y::Real, μ::Real) = -2 * logmxp1(y /...