Small sum up for SG14 feedback

43 views
Skip to first unread message

Matthieu Brucher

unread,
Jan 10, 2019, 5:58:53 PM1/10/19
to SG19 - Machine Learning
Hi all,

I made a small presentation that sums up the discussions we have had here.
I may not be able to attend tomorrow's meeting as my job's network has lots of restrictions, but I think the text is obvious for Jordi, Johan or Franck. Do not hesitate to comment, modify or amend the slides and comments to clarify the picture we want to convey to SG14.

I also added some text to SG14 prior art based on Python (and xtensor as a trial to match Numpy for C++).

Cheers,

Matthieu
--
SG19-LinearAlgebra-01.pptx

Frank Seide

unread,
Jan 10, 2019, 7:46:02 PM1/10/19
to Matthieu Brucher, SG19 - Machine Learning

Thanks! Under “Why do we want nice C++ support”, I would add

 

·         While training can be done in Python, inference is often done in apps that are written in C++

--
You received this message because you are subscribed to the Google Groups "SG19 - Machine Learning" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sg19+uns...@isocpp.org.
To post to this group, send email to sg...@isocpp.org.
Visit this group at https://groups.google.com/a/isocpp.org/group/sg19/.
To view this discussion on the web visit https://groups.google.com/a/isocpp.org/d/msgid/sg19/CAHCaCkKsaO8_X%3Dd0cavGzJwR-jd8%3DYtw0sWBGa_79cZJERfh9w%40mail.gmail.com.

Emad Barsoum

unread,
Jan 10, 2019, 7:52:30 PM1/10/19
to SG19 - Machine Learning, matthieu...@gmail.com, fse...@microsoft.com

True, but there are cases where training in C++ is needed. Such as online training, Reinforcement learning in games..etc. So it isn't only in data processing.

Thanks,
Emad

Michael Wong

unread,
Jan 10, 2019, 9:05:48 PM1/10/19
to Emad Barsoum, SG19 - Machine Learning, matthieu...@gmail.com, Frank Seide
Thanks Matthieu. We will look at this deck for tomorrow.
Off topic, but what is a better start time for you to avoid any network blockage issues?

johan mabille

unread,
Jan 11, 2019, 2:57:36 AM1/11/19
to Michael Wong, Emad Barsoum, SG19 - Machine Learning, Matthieu Brucher, Frank Seide
Thanks Matthieu. In the section "What would be nice but not mandatory" I would add an operator for slicing like numpy. xtensor provides some syntactic sugar  with operator| but I think that having language support for slicing would be better.

Cheers,

Johan

Matthieu Brucher

unread,
Jan 11, 2019, 4:57:58 AM1/11/19
to SG19 - Machine Learning, ebar...@gmail.com, matthieu...@gmail.com, fse...@microsoft.com
Hi Michael,

The best time would be usually 2 or 3 hours later.
If I can't make it, perhaps someone can replace me presenting these slides? 

Cheers,

Matthieu

Matthieu Brucher

unread,
Jan 11, 2019, 5:09:23 AM1/11/19
to SG19 - Machine Learning, fragga...@gmail.com, ebar...@gmail.com, matthieu...@gmail.com, fse...@microsoft.com
Updated slides with all the comments + some typo fixes
SG19-LinearAlgebra-01.pptx
Reply all
Reply to author
Forward
0 new messages