GitHub user feynmanliang opened a pull request:
https://github.com/apache/spark/pull/7575
Edit [SPARK-8536][MLlib]Generalize OnlineLDAOptimizer to asymmetric
document-topic Dirichlet priors
Modify `LDA` to take asymmetric document-topic prior distributions and
`OnlineLDAOptimizer` to use the asymmetric prior during variational inference.
This PR only generalizes `OnlineLDAOptimizer` and the associated
`LocalLDAModel`; `EMLDAOptimizer` and `DistributedLDAModel` still only support
symmetric `alpha` (checked during `EMLDAOptimizer.initialize`).
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/feynmanliang/spark
SPARK-8536-LDA-asymmetric-priors
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/spark/pull/7575.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #7575
----
commit d4284fa14bdfd9d7f15e98d4600de2e2f7d822d2
Author: Feynman Liang <[email protected]>
Date: 2015-07-09T01:31:24Z
Generalize OnlineLDA to asymmetric priors, no tests
commit 72038ff88e7aad5798cae815530ec4e6217938c2
Author: Feynman Liang <[email protected]>
Date: 2015-07-09T23:04:01Z
Add tests referenced against gensim
commit a6dcf706ac6f43b746d55e369bae098a66ad414e
Author: Feynman Liang <[email protected]>
Date: 2015-07-10T02:31:15Z
Change docConcentration interface and move LDAOptimizer validation to
initialize, add sad path tests
commit 58f1d7b6f2632e00451cf4becf2486cbc9650d03
Author: Feynman Liang <[email protected]>
Date: 2015-07-21T18:16:31Z
Fix from review feedback
----
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
