Okon Samuel

> If so, this is potentially an issue already with the existing API, right? I guess so.

> Otherwise, we would have to do something more drastic, such as arranging the machine constructor to make a deep copy of data provided the constructor by default (with option...

> * `MMI.matrix(X)` is (I think) just a view of the data (someone needs to confirm this), Okay i looked up code at Tables,jl and found out that this is...

@ablaom Have you tried testing this with the previous implementation.

Am still trying to reproduce this issue on my linux PC ```julia (@v1.5) pkg> st Status `~/.julia/environments/v1.5/Project.toml` [6e4b80f9] BenchmarkTools v0.5.0 [13f3f980] CairoMakie v0.3.3 [a93c6f00] DataFrames v0.21.8 [da1fdf0e] FreqTables v0.4.1 [c91e804a]...

Am happy were finally discussing about this. This may reduce package testing time for 3rd party developers.

> MachineLearningDistributions.jl Why don't we just call it `MLDistributions` for short

It appears LIBSVM isn't thread safe https://github.com/JuliaML/LIBSVM.jl/issues/60

The updated docs [here](https://docs.julialang.org/en/v1.6-dev/manual/multi-threading/) can supplement that. Basically the key point is that should avoid functions with side effects (i.e mutates the global state)

lol. @tlienart. It's probably better if we make the sklearn version the default then add a `fast` or `speedup` keyword for the faster version