The real problem with the documentation is that the name of the bug is "single pass algorithm" and yet the paper referenced implies that statistics will have to be accumulated for each iteration of the coordinate descent algorithm. This implies one map-reduce per step of the algorithm which is many more than 1.
We really need a complete description of the algorithm as it is in the code and some justification of how it works. On Wed, Jul 31, 2013 at 10:30 PM, Michael Kun Yang <[email protected]>wrote: > https://issues.apache.org/jira/browse/MAHOUT-1273 > > I upload some docs in issue page which include some examples > But these docs seems not explain the ideas clearly. I probably need to > polish them. > > Feedbacks are very welcome > > On Wednesday, July 31, 2013, Timothy Mann wrote: > > > What is the link to the project page with the algorithm details? > > > > -Timothy > > On Aug 1, 2013 2:03 AM, "Kun Yang" <[email protected]> wrote: > > > > > Thanks for your feedback. > > > > > > Besides the code, I have also upload the details of the algorithm in > the > > > project page. Please take a look. > > > > > > I attached the documents in the email as well. > > > > > > Best > > > > > > ----- Original Message ----- > > > From: "Ted Dunning" <[email protected]> > > > To: "Sebastian Schelter" <[email protected]>, "Ted Dunning" < > > > [email protected]> > > > Cc: "Kun Yang" <[email protected]>, "mahout" <[email protected] > > > > > Sent: Wednesday, July 31, 2013 3:57:49 PM > > > Subject: Re: Review Request 13172: #MAHOUT-1273: Single Pass Algorithm > > for > > > Penalized Linear Regression with Cross Validation on MapReduce > > > > > > > > > This is an automatically generated e-mail. To reply, visit: > > > https://reviews.apache.org/r/13172/ > > > In the paper associated with this code, there is the following magical > > step > > > > > > 19: train the model (2) with train data and calculate the mean > > > squared prediction error pi for test data > > > > > > Until I see a specification of what this algorithm is supposed to be > > > doing, I don't see the point in reviewing the code. > > > > > > > > > - Ted Dunning > > > > > > > > > On July 31st, 2013, 10:56 p.m. UTC, Kun Yang wrote: > > > Review request for mahout, Sebastian Schelter and Ted Dunning. > > > By Kun Yang. > > > > > > Updated July 31, 2013, 10:56 p.m. > > > Repository: mahout-git Description > > > Single-Pass Penalized Linear Regression with Cross Validation > on > > > MapReduce > > > > > > For details, please go through the documentation. Testing > > > Units tests for the classes Diffs > > > > > > > > > * > > > > > > core/src/main/java/org/apache/mahout/regression/penalizedlinear/PenalizedLinearConstants.java > > > (PRE-CREATION) > > > * > > > > > > core/src/main/java/org/apache/mahout/regression/penalizedlinear/PenalizedLinearDriver.java > > > (PRE-CREATION) > > > * > > > > > > core/src/main/java/org/apache/mahout/regression/penalizedlinear/PenalizedLinearKeySet.java > > > (PRE-CREATION) > > > * > > > > > > core/src/main/java/org/apache/mahout/regression/penalizedlinear/PenalizedLinearMapper.java > > > (PRE-CREATION) > > > * > > > > > > core/src/main/java/org/apache/mahout/regression/penalizedlinear/PenalizedLinearReducer.java > > > (PRE-CREATION) > > > * > > > > > > core/src/main/java/org/apache/mahout/regression/penalizedlinear/PenalizedLinearSolver.java > > > (PRE-CREATION) > > > * > > > > > > core/src/test/java/org/apache/mahout/regression/penalizedlinear/PenalizedLinearTest.java > > > (PRE-CREATION) > > > * > > > > > > examples/src/main/java/org/apache/mahout/regression/penalizedlinear/Job.java > > > (PRE-CREATION) > > > * > > > > > > examples/src/main/java/org/apache/mahout/regression/penalizedlinear/LinearCrossValidation.java > > > (PRE-CREATION) > > > * > > > > > > examples/src/main/java/org/apache/mahout/regression/penalizedlinear/LinearRegularizePath.java > > > (PRE-CREATION) > > > * examples/src/main/resources/prostate.txt (PRE-CREATION) > > > * > > > > > > examples/src/test/java/org/apache/mahout/regression/penalizedlinear/PenalizedLinearExampleTest.java > > > (PRE-CREATION) > > > * > > > > > > integration/src/main/java/org/apache/mahout/regression/feature/extractor/FeatureExtractUtility.java > > > (PRE-CREATION) > > > * > > > > > > integration/src/main/java/org/apache/mahout/regression/feature/extractor/FeatureExtractorKeySet.java > > > (PRE-CREATION) > > > * > > >
