<@U5F0TT0UX> I think it will be interesting for yo...

# science

[Oleg Yukhnevich]

@altavir I think it will be interesting for you https://github.com/Kotlin/kotlin-numpy

[altavir]

We had a long discussion with @roman.belov about it on Kotlinconf. There are some problems with numpy integration and it is not clear if we really want it because of some bad architectural decissions about it. It is a good project from the point of view of Python to kotlin migration, but I am not sure that it should be used inside kotlin ecosystem.

[altavir]

I think we should discussit more.

[thomasnield]

I'm actually fond of what @altavir did with kmath. Yeah I'm sure it needs to be scaled and hashed out to be a global library, but I think the concept itself is quite solid.

[thomasnield]

Sorry my engagement has been sparse. My time is getting dominated by two apps I'm building and trying to bring to market, some O'Reilly projects (stuck with Python), and teaching some classes at University of Southern California next year.

[thomasnield]

One of the apps on the back burner NEEDS a linear algebra system though. I have a patent on the idea but it's going to require significant investment to build, and I'm planning on making it in Kotlin. Hoping WebAssembly matures enough too so it can be web-based.

[altavir]

kmath is about API, not about performance, but I think what we now have in kmath shows that it is possible to add kmath bindings to high-performant libraries without sacrificing performance. Last addition is Viktor: https://github.com/mipt-npm/kmath/tree/dev/kmath-viktor. The problem with numpy is that at actually requires python interpreter to run and it will be very hard to make stable and reproducable bindings.

[altavir]

@thomasnield kmath is multiplatform and it is possible to use it for linear algebra. My tests for my multiplatform implementations show that they are slower than specialized libraries, but not order of magnitude slower. I am sure I can add whatever you need quite fast if you could formulate what you actually need. Core linear algebra is alteady there.

[thomasnield]

@altavir That's what I was insinuating. I may be helping contribute to it later.

[altavir]

I have a team to work on it. The main contribution would be to state what do you actually need 🙂. I do not use linear algebra much, so it hard for me to understand.

[thomasnield]

@altavir oh yeah I forgot you had resources. Neat! Thanks I will keep that in mind. I have some Markov/Bayesian stuff I'm playing with early next year so I'll let you know.

[altavir]

We will start on SImBa (Simulation with Bayesian tools) next year, so some tools will be complimentary, I hope. I also want to create bridge for rather powerful https://github.com/bat/BAT.jl

[thomasnield]

I need to still get Kotlin-Statistics multplatform (which I believe you started) and using kmath.

[altavir]

We need to check, but I believe I've finished the multiplatform completion, did I forget to create a pull request?..

[thomasnield]

@altavir you did put in the interfaces if I recall, but I need to look at the implementations too

[altavir]

https://github.com/thomasnield/kotlin-statistics/tree/mpp-dev

[altavir]

It should be working. If not, I will remedy that ASAP

[thomasnield]

Oh you put in a PR

[thomasnield]

I did not realize there was a PR. I missed that.

[altavir]

It is in a separate branch

[thomasnield]

Haha, I was getting married at the time so that might be why I missed it. Sorry about that. I will download and build this week

[altavir]

Getting married is more important, I can tell you

[altavir]

did it properly only the second time

[thomasnield]

Thanks again for doing this, that's huge.