reef-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Taegeon Um (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (REEF-1663) Fix TestKMeansOnDirectRunViaFileSystem failure in AppVeyor
Date Wed, 02 Aug 2017 02:04:00 GMT

     [ https://issues.apache.org/jira/browse/REEF-1663?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Taegeon Um updated REEF-1663:
-----------------------------
    Fix Version/s:     (was: 0.16)

> Fix TestKMeansOnDirectRunViaFileSystem failure in AppVeyor
> ----------------------------------------------------------
>
>                 Key: REEF-1663
>                 URL: https://issues.apache.org/jira/browse/REEF-1663
>             Project: REEF
>          Issue Type: Sub-task
>          Components: REEF.NET
>            Reporter: Mariia Mykhailova
>            Assignee: Mariia Mykhailova
>            Priority: Minor
>
> {{O.A.R.Tests.Functional.ML.KMeans.TestKMeans.TestKMeansOnDirectRunViaFileSystem}} [failed
on AppVeyor|https://ci.appveyor.com/project/ApacheSoftwareFoundation/reef/build/1124-master/job/j85983ju7qyytgdc/tests]
with error
> {noformat}
> System.InvalidOperationException : The new loss 14658.95 is larger than previous loss
14658.95, while loss function must be monotonically decreasing across iterations
> {noformat}
> I doubt that the new loss was really significantly less than the previous one, since
the displayed values are identical. But the test uses exact comparisons of floating-point
numbers, which occasionally might produce unexpected results.
> We should probably replace exact comparisons with a proper comparison function, as described
[here|http://stackoverflow.com/questions/3874627/floating-point-comparison-functions-for-c-sharp].
Note that this comparison is done in two files - {{TestKMeans.cs}} and {{KMeansMasterTask.cs}}.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Mime
View raw message