You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@systemml.apache.org by "Niketan Pansare (JIRA)" <ji...@apache.org> on 2016/10/01 20:15:20 UTC
[jira] [Comment Edited] (SYSTEMML-1000) Allow users to pass non-1
bias filler in conv_builtin.dml
[ https://issues.apache.org/jira/browse/SYSTEMML-1000?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15539093#comment-15539093 ]
Niketan Pansare edited comment on SYSTEMML-1000 at 10/1/16 8:14 PM:
--------------------------------------------------------------------
It would be good to support different bias fillers: https://github.com/niketanpansare/incubator-systemml/blob/d9e6efaf297b1a22fcbe3eb0b7f75f07e19969db/src/main/java/org/apache/sysml/api/dl/utils/FillerUtils.java#L31
- constant: requires weightFiller.getValue()
- uniform: requires weightFiller.getMin() and weightFiller.getMax()
- gaussian: requires weightFiller.getMean() and weightFiller.getStd()
- xavier: See [Bengio and Glorot 2010]: Understanding the difficulty of training deep feedforward neuralnetworks
Implemented in https://github.com/niketanpansare/incubator-systemml/blob/d9e6efaf297b1a22fcbe3eb0b7f75f07e19969db/src/main/java/org/apache/sysml/api/dl/utils/FillerUtils.java#L98
- msra: See [Saxe, McClelland, and Ganguli 2013 (v3)].
Implemented in https://github.com/niketanpansare/incubator-systemml/blob/d9e6efaf297b1a22fcbe3eb0b7f75f07e19969db/src/main/java/org/apache/sysml/api/dl/utils/FillerUtils.java#L124
was (Author: niketanpansare):
It would be good to support different bias fillers: https://github.com/niketanpansare/incubator-systemml/blob/d9e6efaf297b1a22fcbe3eb0b7f75f07e19969db/src/main/java/org/apache/sysml/api/dl/utils/FillerUtils.java#L31
> Allow users to pass non-1 bias filler in conv_builtin.dml
> ---------------------------------------------------------
>
> Key: SYSTEMML-1000
> URL: https://issues.apache.org/jira/browse/SYSTEMML-1000
> Project: SystemML
> Issue Type: Wish
> Reporter: Niketan Pansare
>
> Useful pre-step to incorporate nn functions with caffe
> https://github.com/apache/incubator-systemml/blob/master/scripts/staging/SystemML-NN/nn/layers/conv_builtin.dml#L72
> https://github.com/apache/incubator-systemml/pull/158
> [~mwdusenb@us.ibm.com]
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)