[
https://issues.apache.org/jira/browse/HAMA-233?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sangwon Seo updated HAMA-233:
-----------------------------
Attachment: HAMA-233_v01.patch
This is the first patch for CG method.
It is implemented in SparseMatrix.
That is, Cramer rule is for DenseMatrix, and this i for SparseMatrix.
Surely, both are still have problems.
After testing more, I will upload the second patch soon.
Please do not test this patch until futher patches.
(It is not tested !!!)
> Solving linear solution with Conjugate Gradient Method
> ------------------------------------------------------
>
> Key: HAMA-233
> URL: https://issues.apache.org/jira/browse/HAMA-233
> Project: Hama
> Issue Type: New Feature
> Components: matrix
> Affects Versions: 0.1.0
> Reporter: Sangwon Seo
> Assignee: Sangwon Seo
> Priority: Minor
> Fix For: 0.1.0
>
> Attachments: HAMA-233_v01.patch
>
>
> For spase matrix, CG method is good in a way of iterative learning.
> This can be applied to many applications (MLP, descent gradient search,
> etc..) as a primitive, not even in linear solution.
> Related paper (From Carnegie Mellon, It presents in verry simple way to
> understand everything behind the CG) is as bellow.
> http://smiler82.springnote.com/pages/3827983/attachments/2558573
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
