[GitHub] samskalicky commented on issue #12327: [Feature Request] support of diag for N-d arrays

[GitBox <gi...@apache.org>]

samskalicky commented on issue #12327: [Feature Request] support of diag for N-d arrays
URL: https://github.com/apache/incubator-mxnet/issues/12327#issuecomment-418465329
 
 
   @jasonyu1996 I have been prototyping a general diagonal operator based on the numpy implementation. The issue I see is that numpy uses views, so the diagonal operator simply returns a 'diagonal view' to the original data layout in memory. Where as we want an actual copy of the data in the diagonal operator in mxnet. Seemingly we should be able to just traverse the view to copy the resulting diagonal values. 
   
   Im not a deep learning expert, so I dont understand the need/usage of a diagonal operator in a network. Can you provide some motivation or use-cases for where a diagonal operator might be needed in a model?

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services