--
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/sg19/1224d28e-a0a5-4e69-970f-b35b1010ae78%40isocpp.org.
Could we define dynamic graph (in the NN context)? Do we mean:
Thanks,
Frank
--
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/abd48241-698c-42f6-9edd-bd9f803f8909%40isocpp.org.
Let me add a requirement:
8. Highly efficient
This is necessary for dynamic networks with automatic batching, where one not only creates the graph dynamically for each minibatch, but also runs through each batch item separately. I prototyped this (and auto-batching) with CNTK, and since the CNTK graph library was not optimized for speed (since in default CNTK, the graph was only created once), I found graph-creation cost to be a major bottleneck.
From: Emad Barsoum [mailto:ebar...@gmail.com]
Sent: Monday, March 11, 2019 11:18
To: SG19 - Machine Learning <sg...@isocpp.org>
Subject: Re: Graph
--
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/sg19/0c3d8e60-6229-4b13-a2f0-cfa7048ee94c%40isocpp.org.
That sounds like a good start. I think everyone would be looking for the algorithm-iterator-container abstractions like STL. If those don’t exist then it would be perceived as incomplete.
From: Richard Dosselmann <doss...@cs.uregina.ca>
Sent: Sunday, March 10, 2019 3:19 PM
To: SG19 - Machine Learning <sg...@isocpp.org>
Subject: Re: Graph
I have looked at BGL many times. I have also considered graphs in Python and Java. Most such systems seem to have a few fixed graph structures. Given that there are so many types of graphs (and trees), the thinking at the moment is to create components of graphs, such as nodes and edges, that can be assembled to allow users to create whichever type of graph they need. Of course, modern elements of C++ will be useful in doing this.
--
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/d60156f5-ae78-44d2-ab3a-31c54b5a2400%40isocpp.org.
To view this discussion on the web visit https://groups.google.com/a/isocpp.org/d/msgid/sg19/DM5PR03MB26330CF619EF8E3CF9CBD524DD480%40DM5PR03MB2633.namprd03.prod.outlook.com.