[
https://issues.apache.org/jira/browse/FLINK-1707?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15351211#comment-15351211
]
Josep Rubió commented on FLINK-1707:
------------------------------------
Agree, I put this same statement in the design doc. But the paper does not
explain that running in parallel the calculations of E and I vertices of the
binary model gives different intermediate results as it has a different
scheduling but I think it should give same clusters (but I'm not an expert in
bayesian networks).
About the Capacitated Affinity Propagation, one of the advantages of the binary
model is that adding constraints to clusters means you just need different
calculations for E and I messages. In CAP case only the α(i,j) calculation is
different so this means you just need to modify the functions that updates E
vertices.
By the way, I'm not sure I understand your answer. Do you mean we should work
with the original AP algorithm?
Thanks!
> Add an Affinity Propagation Library Method
> ------------------------------------------
>
> Key: FLINK-1707
> URL: https://issues.apache.org/jira/browse/FLINK-1707
> Project: Flink
> Issue Type: New Feature
> Components: Gelly
> Reporter: Vasia Kalavri
> Assignee: Josep Rubió
> Priority: Minor
> Labels: requires-design-doc
> Attachments: Binary_Affinity_Propagation_in_Flink_design_doc.pdf
>
>
> This issue proposes adding the an implementation of the Affinity Propagation
> algorithm as a Gelly library method and a corresponding example.
> The algorithm is described in paper [1] and a description of a vertex-centric
> implementation can be found is [2].
> [1]: http://www.psi.toronto.edu/affinitypropagation/FreyDueckScience07.pdf
> [2]: http://event.cwi.nl/grades2014/00-ching-slides.pdf
> Design doc:
> https://docs.google.com/document/d/1QULalzPqMVICi8jRVs3S0n39pell2ZVc7RNemz_SGA4/edit?usp=sharing
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
