Fork me on GitHub
#off-topic
<
2018-07-23
>
henrik05:07:03

Gig economy + containerization => You compile your code locally, then a guy comes over on a motorcycle, picks up your computer, and rides it over to Amazon.

😆 4
🚢 8
dominicm06:07:20

@alexmiller I was going to say you're up really late, but your profile says you're in the same TZ as me! On holiday?

meguru17:07:25

Hello I want to study Natual language processing such as seq2seq or Doc2Vec. So I would like to use Clojure, Common Lisp, Hylang. Would you advise of each strength and which one is better?

john20:07:59

yeah, anecdotally, I'd argue Clojure is more mature in the ML space than those other two. Not too familiar with Hylang though.

cvic20:07:36

Hy is an interesting Lisp dialect. But I have no idea how mature is the interaction with Python libs...

john20:07:05

oh, right, it's on python... That actually might be the shortest path, in the short term. I think Clojure, while not as mature as Python there, has the most room for growth though in the ML space.

cvic20:07:31

You can cheat in Clojure too, by directly using mature Java libs like Apache OpenNLP or Stanford CoreNLP

john20:07:26

Right. And Carin Meier recently wrapped mxnet which got picked up by the apache foundation, which is what Amazon has blessed off on for their datacenter

cvic21:07:49

Woah, very cool. Are there any other Clojure projects adopted by the Apache Foundation?

john21:07:35

I believe Nathan Marz's Storm got picked up by Apache after it lived at Twitter for a while

meguru21:07:33

Thanks your research! I saw it using NLTK using Hylang, but probably thought that writing with Python would be more efficient. I think I'll start with using Deeplearning 4j at Clojure.

john21:07:46

Nice, my buddy/colleague actually wrote that wrapper: https://github.com/yetanalytics/dl4clj

👍 8
meguru21:07:10

Thank you, I will refer you.