search

scalanlp / breeze

Breeze is/was a numerical processing library for Scala.
https://www.scalanlp.org
Apache License 2.0
3.45k stars 693 forks source link

Failed test #100

Closed jasonbaldridge closed 10 years ago

jasonbaldridge commented 10 years ago

David - Can you check whether you get this same failed test? (I just updated to latest.)

[info] IntMathTest: [info] - ipow [info] - optimize a simple multivariate gaussian with projection [info] - optimize a simple multivariate gaussian with l2 regularization [info] - variance [info] - mode [info] - mean Expected 9.999899999999998E9 but got 9.076197465412052E9 Expected 9.999899999999998E9 but got 9.079955314848719E9 [info] - variance * FAILED * [info] GeneratorDrivenPropertyCheckFailedException was thrown during property evaluation. info [info] Falsified after 0 successful property evaluations. [info] Location: (MomentsTestBase.scala:38) [info] Occurred when passed generated values ( [info] arg0 = Geometric(1.0E-5) [info] )

dlwh commented 10 years ago

Yeah, Geometric fails sometimes, as does Gamma. I haven't decided whether it's a bug or just randomness biting, so I haven't fixed the random seed, as a kind of reminder to myself. I think Gamma is probably actually broken, but I've never gotten around to checking.

Geometric is probably just random chance.

-- David

On Sat, Nov 2, 2013 at 3:17 PM, Jason Baldridge notifications@github.comwrote:

David - Can you check whether you get this same failed test? (I just updated to latest.)

[info] IntMathTest: [info] - ipow [info] - optimize a simple multivariate gaussian with projection [info] - optimize a simple multivariate gaussian with l2 regularization [info] - variance [info] - mode [info] - mean Expected 9.999899999999998E9 but got 9.076197465412052E9 Expected 9.999899999999998E9 but got 9.079955314848719E9 [info] - variance * FAILED * [info] GeneratorDrivenPropertyCheckFailedException was thrown during property evaluation. info [info] Falsified after 0 successful property evaluations. [info] Location: (MomentsTestBase.scala:38) [info] Occurred when passed generated values ( [info] arg0 = Geometric(1.0E-5) [info] )

— Reply to this email directly or view it on GitHubhttps://github.com/scalanlp/breeze/issues/100 .

jasonbaldridge commented 10 years ago

Okay. Maybe makes sense to provide some tolerance for that test.

On Sat, Nov 2, 2013 at 6:55 PM, David Hall notifications@github.com wrote:

Yeah, Geometric fails sometimes, as does Gamma. I haven't decided whether it's a bug or just randomness biting, so I haven't fixed the random seed, as a kind of reminder to myself. I think Gamma is probably actually broken, but I've never gotten around to checking.

Geometric is probably just random chance.

-- David

On Sat, Nov 2, 2013 at 3:17 PM, Jason Baldridge notifications@github.comwrote:

David - Can you check whether you get this same failed test? (I just updated to latest.)

[info] IntMathTest: [info] - ipow [info] - optimize a simple multivariate gaussian with projection [info] - optimize a simple multivariate gaussian with l2 regularization [info] - variance [info] - mode [info] - mean Expected 9.999899999999998E9 but got 9.076197465412052E9 Expected 9.999899999999998E9 but got 9.079955314848719E9 [info] - variance * FAILED * [info] GeneratorDrivenPropertyCheckFailedException was thrown during property evaluation. info [info] Falsified after 0 successful property evaluations. [info] Location: (MomentsTestBase.scala:38) [info] Occurred when passed generated values ( [info] arg0 = Geometric(1.0E-5) [info] )

— Reply to this email directly or view it on GitHub< https://github.com/scalanlp/breeze/issues/100> .

— Reply to this email directly or view it on GitHubhttps://github.com/scalanlp/breeze/issues/100#issuecomment-27635518 .

Jason Baldridge Associate Professor, Dept. of Linguistics, UT Austin Co-founder & Chief Scientist, People Pattern http://www.jasonbaldridge.com http://twitter.com/jasonbaldridge

dlwh commented 10 years ago

There is lots of tolerance: the values can be off by 5%, there's 10k samples, and it gets two tries to get it right for each parameter value tested... which makes me think it's a bug. (Also the fact that when it's wrong both times, the estimated moments generated are typically about the same value...)

-- David

On Sat, Nov 2, 2013 at 8:02 PM, Jason Baldridge notifications@github.comwrote:

Okay. Maybe makes sense to provide some tolerance for that test.

On Sat, Nov 2, 2013 at 6:55 PM, David Hall notifications@github.com wrote:

Yeah, Geometric fails sometimes, as does Gamma. I haven't decided whether it's a bug or just randomness biting, so I haven't fixed the random seed, as a kind of reminder to myself. I think Gamma is probably actually broken, but I've never gotten around to checking.

Geometric is probably just random chance.

-- David

On Sat, Nov 2, 2013 at 3:17 PM, Jason Baldridge < notifications@github.com>wrote:

David - Can you check whether you get this same failed test? (I just updated to latest.)

[info] IntMathTest: [info] - ipow [info] - optimize a simple multivariate gaussian with projection [info] - optimize a simple multivariate gaussian with l2 regularization [info] - variance [info] - mode [info] - mean Expected 9.999899999999998E9 but got 9.076197465412052E9 Expected 9.999899999999998E9 but got 9.079955314848719E9 [info] - variance * FAILED * [info] GeneratorDrivenPropertyCheckFailedException was thrown during property evaluation. info [info] Falsified after 0 successful property evaluations. [info] Location: (MomentsTestBase.scala:38) [info] Occurred when passed generated values ( [info] arg0 = Geometric(1.0E-5) [info] )

— Reply to this email directly or view it on GitHub< https://github.com/scalanlp/breeze/issues/100> .

— Reply to this email directly or view it on GitHub< https://github.com/scalanlp/breeze/issues/100#issuecomment-27635518> .

Jason Baldridge Associate Professor, Dept. of Linguistics, UT Austin Co-founder & Chief Scientist, People Pattern http://www.jasonbaldridge.com http://twitter.com/jasonbaldridge

— Reply to this email directly or view it on GitHubhttps://github.com/scalanlp/breeze/issues/100#issuecomment-27637816 .

jasonbaldridge commented 10 years ago

Okay -- sounds like a candidate for an issue if there isn't one already!

On Sat, Nov 2, 2013 at 10:09 PM, David Hall notifications@github.comwrote:

There is lots of tolerance: the values can be off by 5%, there's 10k samples, and it gets two tries to get it right for each parameter value tested... which makes me think it's a bug. (Also the fact that when it's wrong both times, the estimated moments generated are typically about the same value...)

-- David

On Sat, Nov 2, 2013 at 8:02 PM, Jason Baldridge notifications@github.comwrote:

Okay. Maybe makes sense to provide some tolerance for that test.

On Sat, Nov 2, 2013 at 6:55 PM, David Hall notifications@github.com wrote:

Yeah, Geometric fails sometimes, as does Gamma. I haven't decided whether it's a bug or just randomness biting, so I haven't fixed the random seed, as a kind of reminder to myself. I think Gamma is probably actually broken, but I've never gotten around to checking.

Geometric is probably just random chance.

-- David

On Sat, Nov 2, 2013 at 3:17 PM, Jason Baldridge < notifications@github.com>wrote:

David - Can you check whether you get this same failed test? (I just updated to latest.)

[info] IntMathTest: [info] - ipow [info] - optimize a simple multivariate gaussian with projection [info] - optimize a simple multivariate gaussian with l2 regularization [info] - variance [info] - mode [info] - mean Expected 9.999899999999998E9 but got 9.076197465412052E9 Expected 9.999899999999998E9 but got 9.079955314848719E9 [info] - variance * FAILED * [info] GeneratorDrivenPropertyCheckFailedException was thrown during property evaluation. info [info] Falsified after 0 successful property evaluations. [info] Location: (MomentsTestBase.scala:38) [info] Occurred when passed generated values ( [info] arg0 = Geometric(1.0E-5) [info] )

— Reply to this email directly or view it on GitHub< https://github.com/scalanlp/breeze/issues/100> .

— Reply to this email directly or view it on GitHub< https://github.com/scalanlp/breeze/issues/100#issuecomment-27635518> .

Jason Baldridge Associate Professor, Dept. of Linguistics, UT Austin Co-founder & Chief Scientist, People Pattern http://www.jasonbaldridge.com http://twitter.com/jasonbaldridge

— Reply to this email directly or view it on GitHub< https://github.com/scalanlp/breeze/issues/100#issuecomment-27637816> .

— Reply to this email directly or view it on GitHubhttps://github.com/scalanlp/breeze/issues/100#issuecomment-27637882 .

Jason Baldridge Associate Professor, Dept. of Linguistics, UT Austin Co-founder & Chief Scientist, People Pattern http://www.jasonbaldridge.com http://twitter.com/jasonbaldridge