Carlos Parada
Carlos Parada
> There is no automatic optimisation happening at all. Yep, that would be the issue I'm raising--I'd like it if there was. > Since you mentioned JAX, the Mojo code...
Related to #1255 in general. Mojo has a lot of difficult-to-understand keywords/names/syntax (and in some cases, even the semantics are unintuitive, e.g. Mojo is weakly typed unlike Python, as seen...
I see, thanks. In that case, maybe this should be documented, and the error message replaced to be clearer?
> `Zip` expects `Transducer`s in its arguments and gives a new combined transducer. So the use case is something like this OK, so in that case, it probably makes sense...
Thank you so much for opening this issue! Would you be willing to open a PR implementing the more precise algorithms? (Is there any performance cost to doing so?)
That would be great, thank you! It might also be worth checking out what OnlineStats.jl does. IME it usually has cleaner/faster/more accurate implementations than we do.
Thanks, sorry about that.
I'll note that this can be done automatically using the lazy computation provided by something like Transducers.jl, or graph reduction algorithms in general.
> I originally intended to create a PR right away, but I'm a bit unsure about how this combined with the in-built multidimensionality of the histograms. Steps are one-dimensional in...
BTW, @devmotion, the reason why I'm interested in Transducers.jl is I'm working on a PR that fixes all of the loops and uses of `@inbounds` in StatsBase.jl; Transducers.jl can replace...