[jira] [Created] (SPARK-2273) online learning algorithms: Passive Aggressive

["Jun Xie (JIRA)" <ji...@apache.org>]

Jun Xie created SPARK-2273:
------------------------------

             Summary: online learning algorithms: Passive Aggressive
                 Key: SPARK-2273
                 URL: https://issues.apache.org/jira/browse/SPARK-2273
             Project: Spark
          Issue Type: New Feature
          Components: MLlib
    Affects Versions: 1.0.0
         Environment: OS: Fedora Linux
Code Version: git clone of 1.0.0 version of spark
            Reporter: Jun Xie


Margin-based online learning algorithms can be used in various prediction tasks, including binary and multiclass categorization, regression, uniclass prediction and sequence prediction. The update steps of different algorithms can be unified to a simple constraint optimization problem. 

Koby Crammer (2006) proposed online passive-aggressive algorithm to optimize the proposed constraint problem. It can be used in large-scale problem and maintain a high accuracy of the classifiers by a proved smallest loss which can be attained by any fixed hypothesis.

Reference:
Online Passive-Aggressive Algorithms. Koby Crammer, Ofer Dekel, Joseph Keshet, Shai Shalev-Shwartz, Yoram Singer. JMLR. 7(Mar):551--585, 2006.



--
This message was sent by Atlassian JIRA
(v6.2#6252)