Fork me on GitHub
#data-science
<
2017-10-14
>
gigasquid19:10:27

Really enjoyed getting to meet and talk with everyone that was at the conj

gigasquid19:10:03

Wanted to pass on some notes that I took at the Data Science “Unsession” wish list of things

gigasquid19:10:46

1) A community where we could share information a collaborate more easily (maybe discourse - maybe google groups). A webpage where people could easily find us and libraries would be nice too (like http://www.datahaskell.org/)

gigasquid19:10:16

2) A way for a Jupyter repl to use Clojure easily (Clojure kernel?)

sparkofreason15:10:01

You can use Clojure in Jupyter via BeakerX: https://github.com/twosigma/beakerx

sparkofreason15:10:50

BeakerX also has a pretty decent viz library in Java, and they are working on "autotranslation" of data across kernels.

gigasquid19:10:56

3) A way for Clojure to consume Python libs (insert a pony here)

gigasquid19:10:18

4) A clojure library for auto differentiation

gigasquid19:10:13

I was really happy to see all the enthusiasm and energy about data science at the conj. Already I saw people having conversations and joining forces ❤️ !

blueberry19:10:05

@gigasquid wrt the community channel: I think we should stick to an existing channel such as the main Clojure google group (mailing list). There are not enough people active so if we create a separate place I'm afraid it'll be another ghost town. In addition to that, if we continue in one of the existing groups, there will be exposure that will otherwise not happen; many people are kind of interested, and they need to see there is enough activity to be attracted to learn/try something.

blueberry19:10:54

I'm also not against any other suggestion, but I think we should be where the Clojure community is. Slack is imho an unfortunate choice, but most people are here...

nooga20:10:58

Immediately after watching @gigasquid’s talk I thought about pixie-lang being almost Clojure that is very close to python.

nooga20:10:22

I guess pixie-vm could outright call python if ran as an interpreter. Not so sure about pypy generated JIT though.

aaelony21:10:39

I agree with @blueberry. Perhaps also https://www.clojure-toolbox.com/ can be updated with a section for applicable DS libraries(??)

aaelony21:10:29

another idea (perhaps terrible) would be a Clojure library that would translate Clojure to hylang (https://github.com/hylang/hy) macro-style, then slurp the output back into clojure collections.

blueberry22:10:12

Just to drop a notification that I've just added an OpenCL GPU implementation of 50 recently introduced vectorized math functions to Neanderthal. Some are thousands of times faster than java.Math equivalents. https://github.com/uncomplicate/neanderthal/blob/master/src/clojure/uncomplicate/neanderthal/vect_math.clj

blueberry22:10:32

CUDA implementation comes soon to Neanderthal's CUDA engine.

blueberry22:10:39

Easy to use: (pow a b) or (pow! a b c), (exp! a) and so on...