[ https://issues.apache.org/jira/browse/MAHOUT-121?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12740527#action_12740527 ]
Sean Owen commented on MAHOUT-121: ---------------------------------- On the point of variable declaration -- the location of the declaration has no impact on performance. It is not somehow 'declared' multiple times if within a loop. Pulling it out of the loop could even be very slightly slower, if it means you are forced to assign it a dummy value at initialization which is never used. There's one more benefit when referencing objects: an Object reference that exists, say, in a loop -- the reference goes out of scope when the loop finishes and its referent is immediately GC-able. If it's declared outside the loop, the reference exists until the larger containing block finishes. Unless you set it to null, but, that's more maintenance. It's a waste if the referent is large and not actually used anymore. This has actually bitten me in the past. So in general, in Java. declare variables as deep and late as possible. Agree with Shashikant that while using double instead of Double is nice, using -1 (really, use -1.0 -- double literals belong with double values) as a signal value probably isn't right here. Just say < 0.0. > Speed up distance calculations for sparse vectors > ------------------------------------------------- > > Key: MAHOUT-121 > URL: https://issues.apache.org/jira/browse/MAHOUT-121 > Project: Mahout > Issue Type: Improvement > Components: Matrix > Reporter: Shashikant Kore > Assignee: Grant Ingersoll > Attachments: Canopy_Wiki_1000-2009-06-24.snapshot, doc-vector-4k, > MAHOUT-121-cluster-distance.patch, MAHOUT-121-distance-optimization.patch, > MAHOUT-121-new-distance-optimization.patch, mahout-121.patch, > MAHOUT-121.patch, MAHOUT-121.patch, MAHOUT-121.patch, MAHOUT-121.patch, > MAHOUT-121.patch, mahout-121.patch, MAHOUT-121jfe.patch, Mahout1211.patch > > > From my mail to the Mahout mailing list. > I am working on clustering a dataset which has thousands of sparse vectors. > The complete dataset has few tens of thousands of feature items but each > vector has only couple of hundred feature items. For this, there is an > optimization in distance calculation, a link to which I found the archives of > Mahout mailing list. > http://lingpipe-blog.com/2009/03/12/speeding-up-k-means-clustering-algebra-sparse-vectors/ > I tried out this optimization. The test setup had 2000 document vectors > with few hundred items. I ran canopy generation with Euclidean distance and > t1, t2 values as 250 and 200. > > Current Canopy Generation: 28 min 15 sec. > Canopy Generation with distance optimization: 1 min 38 sec. > I know by experience that using Integer, Double objects instead of primitives > is computationally expensive. I changed the sparse vector implementation to > used primitive collections by Trove [ > http://trove4j.sourceforge.net/ ]. > Distance optimization with Trove: 59 sec > Current canopy generation with Trove: 21 min 55 sec > To sum, these two optimizations reduced cluster generation time by a 97%. > Currently, I have made the changes for Euclidean Distance, Canopy and KMeans. > > Licensing of Trove seems to be an issue which needs to be addressed. -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.