You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@opennlp.apache.org by "Vinh Khuc (JIRA)" <ji...@apache.org> on 2014/05/14 05:27:14 UTC
[jira] [Commented] (OPENNLP-682) Add comments to the L-BFGS trainer
sample param files
[ https://issues.apache.org/jira/browse/OPENNLP-682?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13997235#comment-13997235 ]
Vinh Khuc commented on OPENNLP-682:
-----------------------------------
I've added comments for the training parameters. Instead of creating multiple param files, I added the comments into the existing sample param file MaxentQnExperimentalTrainerParams.txt.
> Add comments to the L-BFGS trainer sample param files
> -----------------------------------------------------
>
> Key: OPENNLP-682
> URL: https://issues.apache.org/jira/browse/OPENNLP-682
> Project: OpenNLP
> Issue Type: Improvement
> Components: Machine Learning
> Reporter: Joern Kottmann
> Assignee: Vinh Khuc
> Priority: Trivial
> Fix For: 1.6.0
>
>
> The L-BFGS sample params file qn-trainer-l1.params should contain a comment explaining the possible parameters.
> Which range should be used for L1? Why is L2 set to zero? Are there other parameters which could be set, but are not used in the sample?
> A similar description should be created for the L2 file. It would also be nice if the L2 file could be renamed to the same naming schema as the L1 (qn-trainer-l1.params) file.
--
This message was sent by Atlassian JIRA
(v6.2#6252)