[jira] [Commented] (SPARK-3255) Faster algorithms for logistic regression

["Henry Lin (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SPARK-3255?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14900114#comment-14900114 ] 

Henry Lin commented on SPARK-3255:
----------------------------------

This paper <a href="http://jmlr.csail.mit.edu/proceedings/papers/v28/gopal13.pdf">here</a> has some insight. The paper experiments with different optimization methods for distributed training, and determines that a Log-concavity bound, discovered by David Blei and John Lafferty (2006), scales the best on large datasets.

> Faster algorithms for logistic regression
> -----------------------------------------
>
>                 Key: SPARK-3255
>                 URL: https://issues.apache.org/jira/browse/SPARK-3255
>             Project: Spark
>          Issue Type: Brainstorming
>          Components: MLlib
>            Reporter: Xiangrui Meng
>
> Logistic regression is perhaps the most widely used classification algorithm in industry. We are looking for faster and scalable algorithms for MLlib. We currently have LogisticRegressionWithLBFGS, and the LIBLINEAR group implemented Spark LIBLINEAR: http://www.csie.ntu.edu.tw/~cjlin/libsvmtools/distributed-liblinear/spark/running_spark_liblinear.html
> Welcome to join the discussion and add more candidates.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org