[
https://issues.apache.org/jira/browse/SPARK-6417?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ehsan Mohyedin Kermani updated SPARK-6417:
------------------------------------------
Comment: was deleted
(was: As I am working on the implementation, I have some serious
concerns/questions.
1) Since I am heavily using distributed matrix abstractions provided in linalg,
I realized it's impossible for me to submit it as a separate package when the
implementation is completed. I don't think it's possible to define new
abstractions like Aaron Staple's TFOCS package, for this task.
2) Adding some functionalities like transforming a local matrix to a
BlockMatrix is necessary, as the underlying matrix is a distributed BlockMatrix
and that's the only distributed matrix data structure with minimal matrix
operations suitable for this task (any suggestions?)
3) Because, there's no literature on LP in mapReduce, my only approach is
primal-dual IPM but with underlying distributed matrix data structures, so can
that be called a distributed LP solver? if so in what degree?
Please let me know your opinion and it'd be great to have someone who is
willing to do the code review and guide me before any submission as I'm
relatively new to Scala. )
> Add Linear Programming algorithm
> ---------------------------------
>
> Key: SPARK-6417
> URL: https://issues.apache.org/jira/browse/SPARK-6417
> Project: Spark
> Issue Type: New Feature
> Components: MLlib
> Reporter: Fan Jiang
> Labels: features
>
> Linear programming is the problem of finding a vector x that minimizes a
> linear function fTx subject to linear constraints:
> minxfTx
> such that one or more of the following hold: A·x ≤ b, Aeq·x = beq, l ≤ x ≤ u.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
