Fork me on GitHub
Slack Archive

This page is not created by, affiliated with, or supported by Slack Technologies, Inc.

2019-06-04

Channels

#data-science
<
2019-06-04
>
otfrom08:06:14

@metasoarous @mrevelle what libraries are people using for the above? ^^

drone13:06:44

In the past, I’ve rolled my own with core.matrix. Looks like there are a couple Clojure wrappers of MALLET for LDA. Unfortunately, Clojure doesn’t have a clear answer, like Python and R do, when it comes to recommending implementations of many existing techniques.

metasoarous19:06:03

Well, I implemented the pca in Polis (https://github.com/pol-is/polisMath), because the routine in Incanter was broken. I did an incremental update solution using power iteration (like old-school the origin Page Rank implementation) using core.matrix. I'm sure you wouldn't call it fast compared to something you might implement with Neanderthal (that would be cool to see!), but it hasn't been a bottleneck for us. As for UMAP (and Python things in general), I'm interested in seeing if you could use @UDRJMEFSN's libpython-clj to wrap the standard Python implementation :thinking_face:

alan08:06:40

@otfrom @metasoarous @mrevelle @val_waeselynck it isn't complete yet, but you might want to take a look at this list (https://github.com/scicloj/scicloj/blob/master/resources/templates/md/posts/2019-05-23-clojure-scientists.md#machine-learning) of libraries. I suggest either fastmath (https://github.com/generateme/fastmath) and/or the http://tech.ml stack (https://github.com/techascent/tech.ml). By the way, keep an eye on https://github.com/scicloj for near future new stuff 😄

drone14:06:41

That’s a nice list. There seems to be a fair bit of duplicate work, I hope that doesn’t mean projects will burn out sooner rather than later.

alan14:06:07

That's one of the reasons why we're doing this. We are also doing online open meetings, you can find the past ones here https://www.youtube.com/channel/UCaoZzhNzq-H7YiQczXKuXuw if you're interested

drone15:06:24

thanks, I was actually on the last one

alan17:06:17

Nice! Sorry, it's hard to remember everyone 😅

drone23:06:00

hah, it’s a good problem to have. hopefully means group is growing

❤️ 8
metasoarous11:06:29

New video out on latest Oz feature: Static site generation with live code reloading! https://www.youtube.com/watch?v=4f3fiWcNt7Y

bananadance 12
metasoarous11:06:19

And of course, an accompanying blog post up on the static site I just deployed using this toolkit. http://metasoarous.com/blog/oz-static-site-generation

metasoarous11:06:16

For usage, issues, etc., please see https://github.com/metasoarous/oz

metasoarous11:06:41

Thanks! Hope you enjoy!

jlmr15:06:10

Hi data-science-wizards! Does anyone know if it is possible to get parinfer running in Jupyter Lab with an IClojure kernel?

Daniel Slutsky05:06:49

@U56R03VNW, parinfer is supported by lein-jupyter: https://github.com/clojupyter/lein-jupyter - should be possible with IClojure too, but my guess is that it is not yet supported.

jlmr06:06:37

Thanks, I tried to get this working with IClojure and Jupyter Lab yesterday but wasn’t succesful. Maybe I’ll try to put something together myself.

👍 4