[
https://issues.apache.org/jira/browse/SPARK-20429?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15991099#comment-15991099
]
Andrew Ray commented on SPARK-20429:
------------------------------------
Can you retest your example with Spark 2.2/master. SPARK-18847 probably fixed
your issue.
> [GRAPHX] Strange results for personalized pagerank if node is involved in a
> cycle
> ---------------------------------------------------------------------------------
>
> Key: SPARK-20429
> URL: https://issues.apache.org/jira/browse/SPARK-20429
> Project: Spark
> Issue Type: Bug
> Components: GraphX
> Affects Versions: 2.1.0
> Reporter: Francesco Elia
> Labels: graphx
>
> I'm trying to run the personalized PageRank implementation of GraphX on a
> simple test graph, which is the following:
> Image: https://i.stack.imgur.com/JDv1l.jpg
> I'm a bit confused on some results that I get when I try to compute the PPR
> for a node that is involved in a cycle. For example, the final output for the
> node 12 is as follows:
> (13, 0.0141)
> (7, 0.0141)
> (19, 0.0153)
> (17, 0.0153)
> (20, 0.0153)
> (11, 0.0391)
> (14, 0.0460)
> (15, 0.0541)
> (16, 0.0541)
> (12, 0.1832)
> I would clearly expect that the node 13 would have a much higher PPR value
> (in fact, I would expect it to be the first one after the starting node
> itself). The problem appears as well with other nodes involved in cycles, for
> example for starting node 13 the node 15 has a very low score. From all the
> testing that I have done it seems that for starting nodes that do not
> participate in a cycle the result is exactly how I expect.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
