This is an automated email from the ASF dual-hosted git repository.
srowen pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/spark.git
The following commit(s) were added to refs/heads/master by this push:
new 72af2c0fbc6 [SPARK-44585][MLLIB] Fix warning condition in MLLib
RankingMetrics ndcgAk
72af2c0fbc6 is described below
commit 72af2c0fbc6673a5e49f1fd6693fe2c90141a84f
Author: Guilhem Vuillier <101632595+guilhem-de...@users.noreply.github.com>
AuthorDate: Fri Jul 28 17:29:47 2023 -0500
[SPARK-44585][MLLIB] Fix warning condition in MLLib RankingMetrics ndcgAk
### What changes were proposed in this pull request?
This PR fixes the condition to raise the following warning in MLLib's
RankingMetrics ndcgAk function: "# of ground truth set and # of relevance value
set should be equal, check input data"
The logic for raising warnings is faulty at the moment: it raises a warning
if the `rel` input is empty and `lab.size` and `rel.size` are not equal.
The logic should be to raise a warning if `rel` input is **not empty** and
`lab.size` and `rel.size` are not equal.
This warning was added in the following PR:
https://github.com/apache/spark/pull/36843
### Why are the changes needed?
With the current logic, RankingMetrics will:
- raise incorrect warning when a user is using it in the "binary" mode
(i.e. no relevance values in the input)
- not raise warning (that could be necessary) when the user is using it in
the "non-binary" model (i.e. with relevance values in the input)
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
No change made to the test suite for RankingMetrics:
https://github.com/uchiiii/spark/blob/a172172329cc78b50f716924f2a344517deb71fc/mllib/src/test/scala/org/apache/spark/mllib/evaluation/RankingMetricsSuite.scala
Closes #42207 from guilhem-depop/patch-1.
Authored-by: Guilhem Vuillier
<101632595+guilhem-de...@users.noreply.github.com>
Signed-off-by: Sean Owen <sro...@gmail.com>
---
.../scala/org/apache/spark/mllib/evaluation/RankingMetrics.scala | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git
a/mllib/src/main/scala/org/apache/spark/mllib/evaluation/RankingMetrics.scala
b/mllib/src/main/scala/org/apache/spark/mllib/evaluation/RankingMetrics.scala
index 37e57736574..a3316d8a8fa 100644
---
a/mllib/src/main/scala/org/apache/spark/mllib/evaluation/RankingMetrics.scala
+++
b/mllib/src/main/scala/org/apache/spark/mllib/evaluation/RankingMetrics.scala
@@ -140,6 +140,9 @@ class RankingMetrics[T: ClassTag] @Since("1.2.0")
(predictionAndLabels: RDD[_ <:
* and the NDCG is obtained by dividing the DCG value on the ground truth
set. In the current
* implementation, the relevance value is binary if the relevance value is
empty.
+ * If the relevance value is not empty but its size doesn't match the ground
truth set size,
+ * a log warning is generated.
+ *
* If a query has an empty ground truth set, zero will be used as ndcg
together with
* a log warning.
*
@@ -157,7 +160,7 @@ class RankingMetrics[T: ClassTag] @Since("1.2.0")
(predictionAndLabels: RDD[_ <:
val useBinary = rel.isEmpty
val labSet = lab.toSet
val relMap = Utils.toMap(lab, rel)
- if (useBinary && lab.size != rel.size) {
+ if (!useBinary && lab.size != rel.size) {
logWarning(
"# of ground truth set and # of relevance value set should be equal,
" +
"check input data")
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@spark.apache.org
For additional commands, e-mail: commits-h...@spark.apache.org
