[
https://issues.apache.org/jira/browse/MAHOUT-600?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12988810#comment-12988810
]
Ted Dunning commented on MAHOUT-600:
------------------------------------
I looked at EPSILON, but it is pretty loose. I wanted a tighter constraint.
My guess is that different architectures have slightly different order of
operations, but why what would cause an relative error > 10^-16 is not
something I could guess at off-hand.
I am still dubious of the definition of logNormalize in any case.
> Reported test failure due to exact floating point comparison
> ------------------------------------------------------------
>
> Key: MAHOUT-600
> URL: https://issues.apache.org/jira/browse/MAHOUT-600
> Project: Mahout
> Issue Type: Bug
> Reporter: Ted Dunning
> Fix For: 0.5
>
> Attachments:
> 0001-MAHOUT-600-Improves-floating-point-comparison-in-log.patch
>
>
> Jeremy Lewi kindly pointed this result from his machine out to us:
> {code}
> Test set: org.apache.mahout.math.VectorTest
> -------------------------------------------------------------------------------
> Tests run: 20, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 0.045 sec
> <<< FAILURE!
> testLogNormalize(org.apache.mahout.math.VectorTest) Time elapsed: 0.002 sec
> <<< FAILURE!
> java.lang.AssertionError:
> expected:<{2:0.5345224838248488,1:0.4235990463273581,0:0.2672612419124244}>
> but was:<{2:0.5345224838248488,1:0.423599046327358,0:0.2672612419124244}>
> at org.junit.Assert.fail(Assert.java:91)
> at org.junit.Assert.failNotEquals(Assert.java:645)
> at org.junit.Assert.assertEquals(Assert.java:126)
> at org.junit.Assert.assertEquals(Assert.java:145)
> at
> org.apache.mahout.math.VectorTest.testLogNormalize(VectorTest.java:509)
> at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
> {code}
> I have a patch that cleans this up. Will attach forthwith.
--
This message is automatically generated by JIRA.
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
