Art

Julia lets you carried away. I need to put a type check for `AbstractNearestNeighbors` there.

I understand your concern about a small user base, but I do not see how it is an obstacle for merging into the master. On the contrary, `thrift` has a...

As `rand` behavior changed in recent julia versions, the rework of RNG handling or a test review is required.

You can try to convert your data to `Float32` to reduce memory footprint.

`infovariation` 👍

It does not make sense to use Integer data input for k-means without initial conversion to floating point format. The mean calculation results into a floating point value, I would...

> Assignment into an array will never change its type, so what happen here is that 0.0 is converted to eltype(cweights). Using .= 0 would have the same effect. Sorry,...

> I think this is an odd restriction. In fact the example given on ScikitLearn's page itself is with integer data... That results in floating point array of centers. >...

I've just re-implemented this functionality in #227 to fix ARI calculations.

Here is another implementation: https://github.com/wildart/ManifoldLearning.jl