flink-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Till Rohrmann (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (FLINK-1729) Assess performance of classification algorithms
Date Fri, 22 Apr 2016 09:56:12 GMT

    [ https://issues.apache.org/jira/browse/FLINK-1729?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15253665#comment-15253665

Till Rohrmann commented on FLINK-1729:

Hi [~hoa@insightdatascience.com],

great to hear that you want to contribute to Flink :-) Nothing has been done for this issue
so far. Thus, you can take it over. I've assigned you this issue.

> Assess performance of classification algorithms
> -----------------------------------------------
>                 Key: FLINK-1729
>                 URL: https://issues.apache.org/jira/browse/FLINK-1729
>             Project: Flink
>          Issue Type: New Feature
>          Components: Machine Learning Library
>            Reporter: Till Rohrmann
>            Assignee: hoa nguyen
>              Labels: ML
> In order to validate Flink's classification algorithms (in terms of performance and accuracy),
we should run them on publicly available classification data sets. This will not only serve
as a proof for the correctness of the implementations but will also show how easy the machine
learning library can be used.
> Bottou [1] published some results for the RCV1 dataset using SVMs for classification.
The SVMs are trained using stochastic gradient descent. Thus, they would be a good comparison
for the CoCoA trained SVMs.
> Some more benchmark results and publicly available data sets ca be found here [2].
> Resources:
> [1] [http://leon.bottou.org/projects/sgd]
> [2] [https://github.com/BIDData/BIDMach/wiki/Benchmarks]

This message was sent by Atlassian JIRA

View raw message