You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Aleksey Zinoviev (JIRA)" <ji...@apache.org> on 2018/07/31 14:19:00 UTC
[jira] [Created] (IGNITE-9145) [ML] Add different strategies to
index labels in StringEncoderTrainer
Aleksey Zinoviev created IGNITE-9145:
----------------------------------------
Summary: [ML] Add different strategies to index labels in StringEncoderTrainer
Key: IGNITE-9145
URL: https://issues.apache.org/jira/browse/IGNITE-9145
Project: Ignite
Issue Type: Improvement
Components: ml
Reporter: Aleksey Zinoviev
Assignee: Aleksey Zinoviev
Fix For: 2.7
The main idea to add a few strategies of indexing: sorting and so on.
Currently it supports only one strategy (most popular with zero and less popular with the max index size).
There are can be a few options
* 'frequencyDesc': descending order by label frequency (most frequent label assigned 0)
* 'frequencyAsc': ascending order by label frequency (least frequent label assigned 0)
* 'alphabetDesc': descending alphabetical order
* 'alphabetAsc': ascending alphabetical order
Please, update the method **transformFrequenciesToEncodingValues and add the strategy as a parameter of trainer.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)