You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@systemml.apache.org by "Matthias Boehm (JIRA)" <ji...@apache.org> on 2018/02/06 01:02:00 UTC

[jira] [Created] (SYSTEMML-2132) Autoencoder over 100K x 10K input w/ unnecessary spark ops

Matthias Boehm created SYSTEMML-2132:
----------------------------------------

             Summary: Autoencoder over 100K x 10K input w/ unnecessary spark ops
                 Key: SYSTEMML-2132
                 URL: https://issues.apache.org/jira/browse/SYSTEMML-2132
             Project: SystemML
          Issue Type: Bug
            Reporter: Matthias Boehm


For a dense 100K x 10K input, the current auto encoder script results in unnecessary spark operations for the initial permutation matrix multiply. After closer inspection this is due to incorrect memory estimates for intermediates which is supposed to be a sparse but returns a dense estimate.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)