Github user yanboliang commented on a diff in the pull request:
https://github.com/apache/spark/pull/19156#discussion_r156564056
--- Diff:
mllib/src/test/scala/org/apache/spark/ml/stat/SummarizerSuite.scala ---
@@ -205,67 +207,21 @@ class SummarizerSuite extends SparkFunSuite with
MLlibTestSparkContext {
}
}
- test("debugging test") {
- val df = denseData(Nil)
- val c = df.col("features")
- val c1 = metrics("mean").summary(c)
- val res = df.select(c1)
- intercept[SparkException] {
- compare(res, Seq.empty)
- }
- }
-
- test("basic error handling") {
- val df = denseData(Nil)
- val c = df.col("features")
- val res = df.select(metrics("mean").summary(c), mean(c))
- intercept[SparkException] {
- compare(res, Seq.empty)
- }
- }
+ testExample("single element", Seq((Vectors.dense(0.0, 1.0, 2.0), 2.0)))
- test("no element, working metrics") {
- val df = denseData(Nil)
- val c = df.col("features")
- val res = df.select(metrics("count").summary(c), count(c))
- compare(res, Seq(Row(0L), 0L))
- }
+ testExample("multiple elements (dense)",
+ Seq(
+ (Vectors.dense(-1.0, 0.0, 6.0), 0.5),
+ (Vectors.dense(3.0, -3.0, 0.0), 2.8),
+ (Vectors.dense(1.0, -3.0, 0.0), 0.0)
+ )
+ )
- val singleElem = Seq(0.0, 1.0, 2.0)
- testExample("single element", Seq(singleElem), ExpectedMetrics(
- mean = singleElem,
- variance = Seq(0.0, 0.0, 0.0),
- count = 1,
- numNonZeros = Seq(0, 1, 1),
- max = singleElem,
- min = singleElem,
- normL1 = singleElem,
- normL2 = singleElem
- ))
-
- testExample("two elements", Seq(Seq(0.0, 1.0, 2.0), Seq(0.0, -1.0,
-2.0)), ExpectedMetrics(
- mean = Seq(0.0, 0.0, 0.0),
- // TODO: I have a doubt about these values, they are not normalized.
- variance = Seq(0.0, 2.0, 8.0),
- count = 2,
- numNonZeros = Seq(0, 2, 2),
- max = Seq(0.0, 1.0, 2.0),
- min = Seq(0.0, -1.0, -2.0),
- normL1 = Seq(0.0, 2.0, 4.0),
- normL2 = Seq(0.0, math.sqrt(2.0), math.sqrt(2.0) * 2.0)
- ))
-
- testExample("dense vector input",
- Seq(Seq(-1.0, 0.0, 6.0), Seq(3.0, -3.0, 0.0)),
--- End diff --
Why do you remove the test against ground true value?
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
