[
https://issues.apache.org/jira/browse/HAMA-107?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12648617#action_12648617
]
Samuel Guo commented on HAMA-107:
---------------------------------
My idea:
During the map phrase,
we will map <row-id, row-vector> to <blockid1, <row-id, row-vector-blockslice1>
>, <blockid2, <row-id, row-vector-blockslice2> >, ... , <blockidn, <row-id,
row-vector-blockslicen> >.
During the reduce phrase,
we can collect <blockidm, list<<row-id,row-vector-blockslicem> >, then sort the
list to get the right sub matrix, and store them.
Any comments?
> Blocking job should be a map/reduce job
> ---------------------------------------
>
> Key: HAMA-107
> URL: https://issues.apache.org/jira/browse/HAMA-107
> Project: Hama
> Issue Type: Improvement
> Components: implementation
> Reporter: Edward J. Yoon
> Fix For: 0.1.0
>
>
> DenseMatrix.blocking() used for build a blocked dense matrix. But, it should
> be work on map/reduce to improve speed.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
