[
https://issues.apache.org/jira/browse/MAHOUT-1365?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14025923#comment-14025923
]
ASF GitHub Bot commented on MAHOUT-1365:
----------------------------------------
GitHub user dlyubimov opened a pull request:
https://github.com/apache/mahout/pull/14
MAHOUT-1365 Implicit feedback ALS-WR
in-core and distributed versions. Per document.
What possibly is lacking at this point
(1) an option to do non-WR version (just simplle implicit feedback).
(2) MSE computations and convergence test.
(3) closer tweaks for in-core math performance. One of the things
side-by-side rbind of U,A and V,A' to mimic vector tuples without getting hands
dirty with actual tuple-filled RDDs. Can it work reasonably well.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/dlyubimov/mahout MAHOUT-1365
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/mahout/pull/14.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 #14
----
commit c47995e74dce76e74bc4a58dbd473120da407dc0
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-05-30T04:32:23Z
ALS
commit 84949a533270cb2bfe735f85321706a424a950cf
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-05-31T00:36:00Z
Bug fixes, test ok
commit 4b525e5011c002bdb4eb37e27699dc1f89f63d63
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-05-31T00:43:31Z
print out iteration rmses in the test
commit 2c9ba249e7dc312c2919a60d43e781eae67c51a9
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-05-31T00:50:36Z
license
commit 565ecb0ca2bb6b1dda30af2c81bdeb1be7fae9fe
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-05-31T01:11:42Z
stopping computation on negative convergence too (i guess in ideal data it
converges too fast and rounding errors could push rmse back up.)
commit 775490da0f28887350c8725909d4af24e8f37310
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-02T22:17:11Z
WIP; bug in zipping -- does several rounds over partitions causing
unidempotant side effects.
commit c92a6cca9eea1c9d15c5fee9c29c79ef9ed3c77f
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-03T00:53:41Z
Fixing bug. Non-idempotent in-place elementwise computations are still an
issue
commit dce371fbdbd17b9c1a3c79260dc8e1180e6073ee
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-03T01:48:33Z
Switching elemnetwise operators to non-inplace execution by default. This
will create GC overhead but it is more benign for side effects.
commit f8ef52fe33dcf6259156ed913ac86dbad6d1a12f
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-03T18:05:52Z
flipping elementwise operator to OK in-place computation by default (back
to where it were). For now. See if the problem surfaces frequent enough to
switch to cloned computations.
commit d9249968cd444ab90d92fa5197a5893f32e8c5a8
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-03T18:42:18Z
Adding regularization parameter. Refactoring decompositions and their tests
into single package (for both distributed and non-distributed stuff).
commit cfa207b1bf98fd941b4d0d48baab4ac35b0deb03
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-03T20:42:12Z
+License on DQR
commit 8e14bf0cf44a5ee86b0c137e8fbb0a8e75bf6ca8
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-03T20:44:01Z
parameter doc
commit fe3d309fd8043e87954b6095040a9e2120668fb8
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-04T01:12:30Z
WIP refreshing Weighed ALS
commit 32d5183543ee032770bfcebe0a5b2e22b3a388e3
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-04T19:31:57Z
removing commented-out section
commit 337a07d8e12b71d79b84c691d6d08348b1713753
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-04T19:33:07Z
Merge branch 'MAHOUT-1566' into MAHOUT-1365
commit 288a21edd01435cd7f1b420513941792d61e19ec
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-05T01:19:52Z
ALS WR in WIP
commit 493f5eabd26c03ec7c8e134f55a217770db481ed
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-05T23:12:18Z
ALS implicit initial write-up
commit 0bf3e483034589c9c9dd1792fea96944a72037f0
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-05T23:48:11Z
scaladoc
commit 5ee0e0d6512628cc10912b71dc2a3002b65adc05
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-06T00:26:07Z
ALS implicit test, fails, bugs.
commit 5e8da2278031a3519c0e708fc9effea567927523
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-09T19:57:13Z
Bug fix
commit 59f9dda0e478b11a46405fea49fe7457eed2eba9
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-09T21:49:08Z
in-memory version for a more transparent prototype
commit 4d877b3b07a83822b01def88e5f77af7cd19a09f
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-09T22:40:08Z
fixes
commit cfc6097e56c5904ea91857f07140b39ec9784da7
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-09T22:53:42Z
Merge branch 'master' into MAHOUT-1365
Conflicts:
math-scala/src/main/scala/org/apache/mahout/math/decompositions/ALS.scala
math-scala/src/main/scala/org/apache/mahout/math/decompositions/package.scala
spark/src/main/scala/org/apache/mahout/sparkbindings/package.scala
spark/src/test/scala/org/apache/mahout/math/decompositions/MathSuite.scala
commit c0a209679c811a97b40649c6afa9399c9d4f20f4
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-06-09T23:01:55Z
fixes
----
> Weighted ALS-WR iterator for Spark
> ----------------------------------
>
> Key: MAHOUT-1365
> URL: https://issues.apache.org/jira/browse/MAHOUT-1365
> Project: Mahout
> Issue Type: Task
> Reporter: Dmitriy Lyubimov
> Assignee: Dmitriy Lyubimov
> Fix For: 1.0
>
> Attachments: distributed-als-with-confidence.pdf
>
>
> Given preference P and confidence C distributed sparse matrices, compute
> ALS-WR solution for implicit feedback (Spark Bagel version).
> Following Hu-Koren-Volynsky method (stripping off any concrete methodology to
> build C matrix), with parameterized test for convergence.
> The computational scheme is following ALS-WR method (which should be slightly
> more efficient for sparser inputs).
> The best performance will be achieved if non-sparse anomalies prefilitered
> (eliminated) (such as an anomalously active user which doesn't represent
> typical user anyway).
> the work is going here
> https://github.com/dlyubimov/mahout-commits/tree/dev-0.9.x-scala. I am
> porting away our (A1) implementation so there are a few issues associated
> with that.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
