[
https://issues.apache.org/jira/browse/SPARK-6425?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14372385#comment-14372385
]
Joseph K. Bradley commented on SPARK-6425:
------------------------------------------
Reinforcement learning is a huge field, and it would be great if Spark were
used for it. I too would like to understand better whether it fits in MLlib.
In particular, I'm wondering:
* Am I correct that the paper you link to is basically using parallel
matrix-vector operations to do RL?
* How common is it to have large enough problems to make distributing this
worthwhile?
> Add parallel Q-learning algorithm to MLLib
> ------------------------------------------
>
> Key: SPARK-6425
> URL: https://issues.apache.org/jira/browse/SPARK-6425
> Project: Spark
> Issue Type: New Feature
> Components: MLlib
> Affects Versions: 1.3.0
> Reporter: zhangyouhua
>
> [~mengxiang]
> Q-learning is a model-free reinforcement learning technique. Specifically,
> Q-learning can be used to find an optimal action-selection policy for any
> given (finite) Markov decision process (MDP). It works by learning an
> action-value function that ultimately gives the expected utility of taking a
> given action in a given state.One of the strengths of Q-learning is that it
> is able to compare the expected utility of the available actions without
> requiring a model of the environment. Additionally, Q-learning can handle
> problems with stochastic transitions and rewards, without requiring any
> adaptations.
> It can be used in artificial intelligence.
> we will use MapReduce for RL with Linear Function Approximation to
> implementation it. some detail can be find
> ：[https://ewrl.files.wordpress.com/2011/08/ewrl2011_submission_11.pdf]
--
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
