[
https://issues.apache.org/jira/browse/GIRAPH-480?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sebastian Schelter updated GIRAPH-480:
--------------------------------------
Description:
I propose to add convergence detection to the RandomWalkVertex. Convergence is
achieved when the overall absolute change (L1 norm) of the difference between
the current and the previous probability vector becomes less than a given
threshold. Convergence detection can be implemented via an additional
aggregator and a check in the master compute function.
This change would make the class much easier to use as the users don't have to
worry about the number of supersteps to execute, but can simply specify a high
number as MAX_SUPERSTEPS and be sure that the algorithm convergences when
acceptable quality of the result is reached.
was:
I propose to add convergence detection to the RandomWalkVertex. Convergence is
achieved when the overall absolute change (L1 norm) of the difference between
the current and the previous probability vector. Convergence detection can be
implemented via an additional aggregator.
This change would make the class much easier to use as the users don't have to
worry about the number of supersteps, but can simply specify a high number as
MAX_SUPERSTEPS and be sure that the algorithm convergences when acceptable
quality of the result is reached.
> Add convergence detection to org.apache.giraph.examples.RandomWalkVertex
> ------------------------------------------------------------------------
>
> Key: GIRAPH-480
> URL: https://issues.apache.org/jira/browse/GIRAPH-480
> Project: Giraph
> Issue Type: Improvement
> Components: examples
> Affects Versions: 0.2.0
> Reporter: Sebastian Schelter
> Assignee: Sebastian Schelter
>
> I propose to add convergence detection to the RandomWalkVertex. Convergence
> is achieved when the overall absolute change (L1 norm) of the difference
> between the current and the previous probability vector becomes less than a
> given threshold. Convergence detection can be implemented via an additional
> aggregator and a check in the master compute function.
> This change would make the class much easier to use as the users don't have
> to worry about the number of supersteps to execute, but can simply specify a
> high number as MAX_SUPERSTEPS and be sure that the algorithm convergences
> when acceptable quality of the result is reached.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
