Fork me on GitHub
#onyx
<
2017-07-08
>
jasonbell08:07:56

@lucasbradstreet @aaelony It’s been on my to do list for a while. I’m interested in the potential of a continual learning model, possibly in Cortex. Time is not on my side at the moment.

matan05:07:57

jasonbell: just curious. why, other than for fun or the feeling of being original (no irony intended), would you use a machine learning framework that virtually no other colleague from a machine learning background would ever collaborate with you through? or do you see something very special in cortex v.s. all the modern frameworks that machine learning people use and collaborate through?

jasonbell07:07:52

@U1YTUBH53 I know quite a few others but some of them are old (Weka etc) on the to do list at the moment are dl4j, Keras and TF. Just I have some exposure to them. And I’m not doing it be original, there are far better Cortex posts out there by @U04VAD6RL for example, far more to learn than I could provide. I am, however, interested to see how the Onyx/ML/DL/AI mix would happen with continually updating models etc. And as for doing things for fun, well I wrote this http://eu.wiley.com/WileyCDA/WileyTitle/productCd-1118889061.html 🙂

matan04:07:32

Nice book!!!

matan14:07:57

Might order from Amazon. Have you actually gotten good performance for machine learning in clojure, compared to the more efficient and popular programming languages you mention in the book?

matan14:07:41

I mean, clojure isn't even used in the clojure compiler because it's slower than Java (sad-but-true) 😂

jasonbell07:07:26

Speeds were fine for most cases. If I had more time I’d redo the examples in the book from Java to Clojure. Might do that when I’m on holiday soon.

lmergen08:07:55

@lucasbradstreet i’m also doing ML on onyx, albeit not yet in production. NLP + neural net + SVM. i have no idea what best practices are in this regard, but i’ll be happy to share with you what worked well for me.