Skip to content

[SPARK-58587][ML] Move final IsotonicRegression PAV pass to executor #57788

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

Closed
zhengruifeng wants to merge 2 commits into from
+12 −4
Closed

[SPARK-58587][ML] Move final IsotonicRegression PAV pass to executor #57788

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f8b621e
test
zhengruifeng Mar 6, 2025
f3ed6ce
[WIP][MLLIB] Use spillable sort for final IsotonicRegression pass
zhengruifeng Aug 5, 2026
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
9 changes: 5 additions & 4 deletions mllib/src/main/scala/org/apache/spark/mllib/regression/IsotonicRegression.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -477,7 +477,8 @@ class IsotonicRegression private (private var isotonic: Boolean) extends Seriali
private def parallelPoolAdjacentViolators(
input: RDD[(Double, Double, Double)]): Array[(Double, Double, Double)] = {
val keyedInput = input.keyBy(_._2)
val parallelStepResult = keyedInput

keyedInput
// Points with same or adjacent features must collocate within the same partition.
.partitionBy(new RangePartitioner(keyedInput.getNumPartitions, keyedInput))
.values
Expand All @@ -486,10 +487,10 @@ class IsotonicRegression private (private var isotonic: Boolean) extends Seriali
// Aggregate points with equal features into a single point.
.map(makeUnique)
.flatMap(poolAdjacentViolators)
// Sort partial results with a spill-capable shuffle before the final PAV pass.
.sortBy(_._2, ascending = true, numPartitions = 1)
.mapPartitions(p => poolAdjacentViolators(p.toArray).iterator)
.collect()
// Sort again because collect() doesn't promise ordering.
.sortBy(_._2)
poolAdjacentViolators(parallelStepResult)
}

/**
Expand Down
7 changes: 7 additions & 0 deletions mllib/src/test/scala/org/apache/spark/mllib/regression/IsotonicRegressionSuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -176,6 +176,13 @@ class IsotonicRegressionSuite extends SparkFunSuite with MLlibTestSparkContext w
assert(model.predictions === Array(1, 2, 3, 4, 5))
}

test("isotonic regression merges partial results across partitions") {
val model = runIsotonicRegressionOnInput(generateIsotonicInput(Seq(1, 3, 2, 4)), true, 2)

assert(model.boundaries === Array(0, 1, 2, 3))
assert(model.predictions === Array(1, 2.5, 2.5, 4))
}

test("weighted isotonic regression") {
val model = runIsotonicRegression(Seq(1, 2, 3, 4, 2), Seq(1, 1, 1, 1, 2), true)

Expand Down