[
https://issues.apache.org/jira/browse/FLINK-3742?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15237686#comment-15237686
]
ASF GitHub Bot commented on FLINK-3742:
---------------------------------------
GitHub user rawkintrevo opened a pull request:
https://github.com/apache/flink/pull/1875
[FLINK-3742] Add Multilayer Perceptron
https://en.wikipedia.org/wiki/Multilayer_perceptron
Multilayer perceptron is a simple sort of artificial neural network. It
creates a directed graph in which the edges are parameter weights and nodes are
non-linear activation functions. It is solved via a method known as back
propagation.
I anticipate this being an ongoing thing for a bit. I'll be working on docs
in the meantime and commit soon.
A note on the tests. Neural-nets take a while to fit by their nature. I
recommend using the warm starts and verifying that it is getting closer after a
few batches of iterations, and some other tests for the unit functionality
(winding/unwinding parameter matrix arrays, etc.)
- [x] General
- The pull request references the related JIRA issue
- The pull request addresses only one issue
- Each commit in the PR has a meaningful commit message
- [ ] Documentation
- Documentation has been added for new functionality
- Old documentation affected by the pull request has been updated
- JavaDoc for public methods has been added
- [ ] Tests & Build
- Functionality added by the pull request is covered by tests
- `mvn clean verify` has been executed successfully locally or a Travis
build has passed
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/rawkintrevo/flink mlp
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/flink/pull/1875.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 #1875
----
commit d8b75aeb5ece02afe35d3c811347c70e18a419b6
Author: Trevor Grant <[email protected]>
Date: 2016-03-28T22:58:28Z
[FLINK][WIP][neural-nets] Initial Commit- working MLP
commit 75167de0b8673c5dd0aa08fd9e5fe96ada8330b4
Author: Trevor Grant <[email protected]>
Date: 2016-03-28T22:58:46Z
[FLINK][WIP][neural-nets] Initial Commit- working MLP
commit fd65d276063ae0dc69233257d8eb9f1e691d19b6
Author: Trevor Grant <[email protected]>
Date: 2016-04-06T22:44:45Z
[FLINK][WIP][neural-nets] Refactored for working PredictoinFunction
commit 81dcac71989e9342c3eb7af31a1a2574a91233c8
Author: Trevor Grant <[email protected]>
Date: 2016-04-08T13:39:59Z
[FLINK][WIP][neural-nets] Cleaned up MLP etc
commit 58116f5433e1d4360b35a92d7151104d5795f2c0
Author: Trevor Grant <[email protected]>
Date: 2016-04-08T15:05:59Z
[FLINK][WIP][neural-nets] Added Warm Starts
commit a88d06fbb9452ee300c990e7dde375738a5a3bb8
Author: Trevor Grant <[email protected]>
Date: 2016-04-12T18:13:41Z
[FLINK][WIP][neural-nets] Clean up Multi-layer Perceptron
----
> Add Multi Layer Perceptron Predictor
> ------------------------------------
>
> Key: FLINK-3742
> URL: https://issues.apache.org/jira/browse/FLINK-3742
> Project: Flink
> Issue Type: Improvement
> Components: Machine Learning Library
> Reporter: Trevor Grant
> Assignee: Trevor Grant
> Priority: Minor
>
> https://en.wikipedia.org/wiki/Multilayer_perceptron
> Multilayer perceptron is a simple sort of artificial neural network. It
> creates a directed graph in which the edges are parameter weights and nodes
> are non-linear activation functions. It is solved via a method known as back
> propagation.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
