A modern programming language that makes developers happier.

kotlinlang

The latest PR brings to `KMath`  finalized optimization and interpolation API as well as unique optimization technique called quasi-optimal weights method (<https://arxiv.org/abs/physics/0604127>). <https://github.com/mipt-npm/kmath/pull/404>. This feature finalizes the feature set for 0.3.0.

By the way, <@U0BA5R9KL> and <@U01K50REX6C>, you could be interested in <https://github.com/mipt-npm/kmath/pull/402>, Iaroslav is in the process of porting Common-Math AD to multiplatform with addition of custom types and algebras. In the end it should allow to do AD computation on arbitrary algebras, including tensors

as far as AD is concerned I think Julia is a place to learn from rather than Common-Math (I am very sceptical of the quality of numerical algorithms in this library). I think we need a hero to port a package like <https://fluxml.ai/Zygote.jl/latest/> to KMP

Julia uses dynamic type substitution, which allows differentiate untyped expressions. Otherwise, I do not think algorithms used there are better. KMath algebras allow to do the same thing without type substitution.

I remember last year an announcement that Facebook was working on AD in kotlin (compiler plugin). Anyone knows anything about that?

They only shown very early syntax prototype in the announce. And nothing after that. It was compiler fork, not a plugin.