Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix flaky ModelInsight tests #395

Merged
merged 7 commits into from
Aug 30, 2019
Merged

Fix flaky ModelInsight tests #395

Changes from 1 commit
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
1084fcf
Update ModelInsightsTest.scala
TuanNguyen27 Aug 27, 2019
ba82754
Update ModelInsightsTest.scala
TuanNguyen27 Aug 28, 2019
5dc762c
Update ModelInsightsTest.scala
TuanNguyen27 Aug 28, 2019
9815b59
Merge branch 'master' into tn/fix_flaky
TuanNguyen27 Aug 28, 2019
72f8f4b
Merge branch 'master' into tn/fix_flaky
tovbinm Aug 29, 2019
5afb446
Update ModelInsightsTest.scala
TuanNguyen27 Aug 30, 2019
b28865c
Update ModelInsightsTest.scala
TuanNguyen27 Aug 30, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 4 additions & 4 deletions core/src/test/scala/com/salesforce/op/ModelInsightsTest.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -755,8 +755,8 @@ class ModelInsightsTest extends FlatSpec with PassengerSparkFixtureTest with Dou
val bigCoeffSum = orginalbigCoeff * math.sqrt(smallFeatureVariance) / labelStd + descaledbigCoeff
val absError2 = math.abs(originalsmallCoeff * math.sqrt(bigFeatureVariance) / labelStd - descaledsmallCoeff)
val smallCoeffSum = originalsmallCoeff * math.sqrt(bigFeatureVariance) / labelStd + descaledsmallCoeff
absError / bigCoeffSum < tol shouldBe true
absError2 / smallCoeffSum < tol shouldBe true
absError / (2 * bigCoeffSum) < tol shouldBe true
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why did you hardcode this change instead of adjusting the tolerance? Is there a reason for a factor of 2 here?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

i think back when i wrote this test you suggested abs(x_1 - x_2) / 2 * (x_1 + x_2) as one way to compute how close x_1 is to x_2, or did I misremember?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I suggested difference / avg. value, which is abs(x_1 - x_2) * 2 / (x_1 + x_2). It probably makes more sense to change it to that, but you should probably increase the overall tolerance too to cut down on flakiness.

absError2 / (2 * smallCoeffSum) < tol shouldBe true
}

it should "correctly return the descaled coefficient for logistic regression, " +
Expand All @@ -771,8 +771,8 @@ class ModelInsightsTest extends FlatSpec with PassengerSparkFixtureTest with Dou
val bigCoeffSum = orginalbigCoeff * math.sqrt(smallFeatureVariance) + descaledbigCoeff
val absError2 = math.abs(originalsmallCoeff * math.sqrt(mediumFeatureVariance) - descaledsmallCoeff)
val smallCoeffSum = originalsmallCoeff * math.sqrt(mediumFeatureVariance) + descaledsmallCoeff
absError / bigCoeffSum < tol shouldBe true
absError2 / smallCoeffSum < tol shouldBe true
absError / (2 * bigCoeffSum) < tol shouldBe true
absError2 / (2 * smallCoeffSum) < tol shouldBe true
}

it should "correctly return moments calculation and cardinality calculation for numeric features" in {
Expand Down