[
https://issues.apache.org/jira/browse/SYSTEMML-686?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15805329#comment-15805329
]
Mike Dusenberry edited comment on SYSTEMML-686 at 1/6/17 7:11 PM:
------------------------------------------------------------------
Copying from GitHub [PR 320 |
https://github.com/apache/incubator-systemml/pull/320]:
[~niketanpansare]:
{quote}
Regarding distributed implementation of convolution and max pooling, there are
following options:
# Current approach (Mini-batch training): for loop \{ CP conv + CP max_pool \}
# Batched approach (applicable for batched training as well as batched
prediction): for loop \{ SPARK conv + SPARK max_pool \}
# Parfor Prediction: remote SPARK parfor loop \{ CP conv + CP max_pool \}
@dusenberrymw Which case do you think we should address first ?
{quote}
was (Author: [email protected]):
Copying from GitHub [PR
320](https://github.com/apache/incubator-systemml/pull/320)
[~niketanpansare]:
{quote}
Regarding distributed implementation of convolution and max pooling, there are
following options:
# Current approach (Mini-batch training): for loop \{ CP conv + CP max_pool \}
# Batched approach (applicable for batched training as well as batched
prediction): for loop \{ SPARK conv + SPARK max_pool \}
# Parfor Prediction: remote SPARK parfor loop \{ CP conv + CP max_pool \}
@dusenberrymw Which case do you think we should address first ?
{quote}
> Implement Spark instructions for convolution and pooling functions
> ------------------------------------------------------------------
>
> Key: SYSTEMML-686
> URL: https://issues.apache.org/jira/browse/SYSTEMML-686
> Project: SystemML
> Issue Type: Task
> Reporter: Niketan Pansare
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
