These are chat archives for beniz/deepdetect

22nd
Oct 2016
Ankush Bhalotia
@ankushb92
Oct 22 2016 02:03
@beniz : Great :) Right now, I'm going through some of the tutorials and trying to get a basic understanding of the system. I have experience mainly in Python. I have familiarity with C++, Java and Javascript as well.
Ankush Bhalotia
@ankushb92
Oct 22 2016 05:13
@beniz: I've been looking at the tickets. Can you suggest a ticket I can take up as a beginner? Any suggestions as to how I should go about understand the codebase?
Emmanuel Benazera
@beniz
Oct 22 2016 05:45
hi @ankushb92 great, one way to get acquainted with the API and capabilities is to write a client. There's a Java client but it's untested, that could be a nice first step. A JS client could be another good step. Digging into the C++ server is less fun and can come later for instance.
Ankush Bhalotia
@ankushb92
Oct 22 2016 05:52
@beniz : Cool. So, for now, I'll look into the Java client and try to contribute there. I'm looking at the open issues there and will try and test it.
Emmanuel Benazera
@beniz
Oct 22 2016 06:53
yes it's been done under my lead but it is untested. You can fork it and try to get it to work, update the doc etc.. don't hesitate if you have questions.
Kumar Shubham
@kyrs
Oct 22 2016 09:28
hi @beniz how are you planning to store the word sequence for LSTM in datum ?
Emmanuel Benazera
@beniz
Oct 22 2016 10:57
I had a plan ;) I took the picture of my notes for you. I ll send it. I think you can store the Delta just as another dimension of the Datum
Kumar Shubham
@kyrs
Oct 22 2016 11:26
thanks !! It would be really helpful.
btw I am a bit confused with present format of storing the words.. I think that present format can't preserve the word ordering..
how are you planning to do that
Emmanuel Benazera
@beniz
Oct 22 2016 15:17
Current storage of words is for BOW, there has been no need for ordered words yet, since we have the support for character level. But we can add it if needed.