[
https://issues.apache.org/jira/browse/FLINK-1745?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15306122#comment-15306122
]
ASF GitHub Bot commented on FLINK-1745:
---------------------------------------
GitHub user danielblazevski opened a pull request:
https://github.com/apache/flink/pull/2048
[Flink 1934] Add approximative k-nearest-neighbours (kNN) algorithm to
machine learning library
I added approximate knn algorithms. In another PR, there are two exact
methods, one basic algorithms using a prirority queue and another using a
quadtree (see: https://github.com/apache/flink/pull/1220 ).
For this PR, I added z-value based knn and LSH (Locality Sensitive Hashing)
based knn. Z-values are good for low-to-moderate dimension. For details, see
the paper [2] someone put on the exact JIRA issue:
https://issues.apache.org/jira/browse/FLINK-1745
https://www.cs.utah.edu/~lifeifei/papers/mrknnj.pdf
As the z-values aren't applicable for larger dimension, so I used -- as the
paper suggests -- a more standard LSH approach.
The paper describes a fairly sophisticated MapReduce (MR) design, which I
did not use. Using the same MR design pattern as the exact method, I found
really good performance improvement! In JIRA, I ran this by @tillrohrmann, and
he was OK with a less optimized version for now. Here is a link for a talk I
recently gave on this, which includes links for the video and slides:
http://www.meetup.com/ny-scala/events/231163636/
Because both the LSH and z-value use the same MR design pattern, I
reformatted the codebase from the PR for exact version a bit to make it more
modular.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/danielblazevski/flink FLINK-1934
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/flink/pull/2048.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 #2048
----
commit c7e5056c6d273f6f0f841f77e0fdd91ca221602d
Author: Chiwan Park <chiwanp...@apache.org>
Date: 2015-06-30T08:41:25Z
[FLINK-1745] [ml] Add exact k-nearest-neighbor join
commit 9d0c7942c09086324fadb29bdce749683a0d1a7e
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-15T21:49:05Z
modified kNN test to familiarize with Flink and KNN.scala
commit 611248e57166dc549f86f805b590dd4e45cb3df5
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-15T21:49:17Z
modified kNN test to familiarize with Flink and KNN.scala
commit 1fd8231ce194b52b5a1bd55bbc5e135b3fa5775b
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-16T01:26:57Z
nightly commit, minor changes: got the filter to work, working on mapping
the training set to include box lables
commit 15d7d2cb308b23e24c43d103b85a76b0e665cbd3
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-22T02:02:51Z
commit before incporporating quadtree
commit 8f2da8a66516565c59df8828de2715b45397cb7f
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-22T15:49:25Z
did a basic import of QuadTree and Test; to-do: modify QuadTree to allow
KNN.scala to make use of
commit e1cef2c5aea65c6f204caeff6348e2778231f98d
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-22T21:03:04Z
transfered ListBuffers for objects in leaf nodes to Vectors
commit c3387ef2ef59734727b56ea652fdb29af957d20b
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-23T00:41:29Z
basic test on 2D unit box seems to work -- need to generalize, e.g. to
include automated bounding box
commit 48294ff37a5f800e5111280da5a3c03f4375028d
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-23T15:03:06Z
had to debug quadtree -- back to testing 2D
commit 6403ba14e240ed8d67a296ac789e7e00dece800d
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-23T15:22:46Z
Testing 2D looks good, strong improvement in run time compared to
brute-force method
commit 426466a40bc2625f390fe0d912f56a346e46c8f8
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-23T19:04:52Z
added automated detection of bounding box based on min/max values of both
training and test sets
commit c35543b828384aa4ce04d56dfcb3d73db46d1e6d
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-24T00:28:56Z
added automated radius about test point to define localized neighborhood,
result runs. TO-DO: Lots of tests
commit 8e2d2e78f8533d4192aebe9b4baa7efbfa5928a5
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-24T00:54:06Z
Note for future: previous commit passed test of Chiwan Park had in intial
knn implementation
commit d6fd40cb88d6e198e52c368e829bf7d32d432081
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-24T01:56:38Z
Note for future: previous commit passed 3D version of the test that Chiwan
Park had in the intial knn implementation
commit 0ec1f4866157ca073341672e7fe9a50871ac0b7c
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-24T14:27:20Z
changed filename of QuadTreeTest to QuadTreeSuite, about to make test more
comprehensive and similar format to other Flink tests
commit ac81561cad27b65d158ae08fd0fb15bdb51d1c8b
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-24T19:51:32Z
refactored testing of QuadTree, and added more tests
commit b17f82d5ce0214617c8dbc4a387410057d6f3832
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-24T22:49:10Z
added KNNBenchmark to check runtimes
commit 530565835d4b5934fcac9e0e51105bb669fec9be
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-24T22:49:17Z
added KNNBenchmark to check runtimes
commit 1f946cb30450604e92bbd0f5959ce9a60eb4c41b
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-25T01:13:00Z
fixed bug -- in find siblings, needed to search for minimal bounding boxes
commit 22e4eb7b57795ad1ca4392ca1c1a8bdae76afa8e
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-27T03:34:16Z
added more thorough benchmark files; about to modify bounding box to only
bound training set and modify search for the testing set
commit 3723f6b09ec7d45f6444df70a5f699cbf998a4bb
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-27T03:34:21Z
added more thorough benchmark files; about to modify bounding box to only
bound training set and modify search for the testing set
commit c41d3e1029bf81a37cf3594f202b904e2d99e3ac
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-28T03:53:12Z
major simplification in choosing the radius to look at nearby neighbors
commit 7c77ea20fd9e8a0c4a33c81b83187c84b380d6b2
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-28T04:03:59Z
cleaned up; commit before deleting previous sibling search
commit cf4aa5d75611db19040466040c3d29432cb0e5f7
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-09-28T14:02:51Z
added new devel branch to push temp changes on github
commit ec6ddb0a57136075b4f77616e6e48eb5bcc50a11
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-10-01T00:17:33Z
cleaned up; removed comments and a unused method
commit 7ed9926d8207b5f59b4ceb968d7ebd732029f5c3
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-10-01T20:28:31Z
fixed bug in bFiltVect, and renamed to trainingFiltered
commit 4b3bb2ec92396bf754d0d207ffc6853406ce7c39
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-10-01T21:05:46Z
fixed bug: if there are fewer than maxPerBox total training points, do not
do heap construction, just make siblingsQueue = root.objects
commit 1662b38822dfdfbbca272d377fa3b94f8246e9e6
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-10-01T22:03:42Z
added metric to constructor, and added a flag to test whether it is
Euclidean or SquaredEuclidean
commit f654b841cc8d91fa861f188469831404c288b227
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-10-01T22:48:51Z
changed name of test that uses the Quadtree along with KNN -- modified from
CHiwan Park's test to ensure flag to use Quadtree will pass
commit 7928798281dba5554eeb63df7e67400a42e7a381
Author: danielblazevski <daniel.blazev...@gmail.com>
Date: 2015-10-01T22:54:25Z
fixed QuadTree test to conform to using a min-heap
----
> Add exact k-nearest-neighbours algorithm to machine learning library
> --------------------------------------------------------------------
>
> Key: FLINK-1745
> URL: https://issues.apache.org/jira/browse/FLINK-1745
> Project: Flink
> Issue Type: New Feature
> Components: Machine Learning Library
> Reporter: Till Rohrmann
> Assignee: Daniel Blazevski
> Labels: ML, Starter
>
> Even though the k-nearest-neighbours (kNN) [1,2] algorithm is quite trivial
> it is still used as a mean to classify data and to do regression. This issue
> focuses on the implementation of an exact kNN (H-BNLJ, H-BRJ) algorithm as
> proposed in [2].
> Could be a starter task.
> Resources:
> [1] [http://en.wikipedia.org/wiki/K-nearest_neighbors_algorithm]
> [2] [https://www.cs.utah.edu/~lifeifei/papers/mrknnj.pdf]
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
