Aleksey Khudyakov

I think proper solution will require changes to linear regression API and consequently bootstrapRegress. I think simply returning NaN is bad idea as this issue shown. Lack of unique solution...

In fact we have two problems here: 1. We need PRNG which could generate multiple independent streams of random numbers from single seed 2. Way to assign these streams to...

Paper itself [2011-Parallel_Random_Numbers:_As_Easy_as_1,_2,_3.pdf](https://github.com/mipt-npm/kmath/files/3268853/2011-Parallel_Random_Numbers._As_Easy_as_1._2._3.pdf)

And of course documentation omits most important part: algorithm being used. Note that period is relatively short 2^64 which could pose problems for MC. As far a I understand extremely...

I'm not sure how faithful kmath about algebra abstractions. But rings don't have division. You'll get field if you want to add division. Also while it's possible to define Field...

I think that main challenge (and main value) is to make things like `length / time` work and produce values of correct type. I know too little about operator overloading...

This approach means that _every_ combination of dimensions should be written explicitly. That's ungodly amount of boilerplate. Take for example: `F::ML/T^2 = m::M · a::L/T^2`. And that's only beginning.

I spent another hour what does `DeserialiseFailure 1 "expect list of length 2"` mean. I decided that enough is enough. I see two possible strategies for fixing this: 1. Modify...

Yes CDDL could be used to define schema for encoding for haskell values. For example data type ```haskell data Foo = Foo { foo :: Int , bar :: Char...

Property `serialize (deserialize x) == x` implies that there's only one way to encode given value. But that's not the case. For example CBOR has two ways of encoding lists....