Ruifeng Zheng created SPARK-60835:
-------------------------------------
Summary: Break label propagation ties with the lowest label
Key: SPARK-60835
URL: https://issues.apache.org/jira/browse/SPARK-60835
Project: Spark
Issue Type: Improvement
Components: GraphX
Affects Versions: 5.0.0
Reporter: Ruifeng Zheng
GraphX label propagation currently resolves equally frequent labels by
selecting the highest label, following SPARK-42856. GraphFrames instead selects
the lowest label to follow the LDBC label propagation specification; Spark
SQL's deterministic mode aggregate also selects the lowest tied value.
Select the lowest label among the labels with maximum frequency. Keep frequency
as the primary criterion and handle the full signed VertexId range without
negating vertex IDs.
For example, if vertex 0 receives labels 1 and 2 once each, select label 1
instead of label 2. Add regression coverage for positive labels, negative
labels, Long.MinValue/Long.MaxValue, and frequency priority.
Review discussion:
https://github.com/apache/spark/pull/60026#issuecomment-6592081340
Related test-only follow-up: https://github.com/apache/spark/pull/60026
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
