[jira] [Commented] (MAHOUT-1059) New matrix extensions

["Ted Dunning (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/MAHOUT-1059?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13447442#comment-13447442 ] 

Ted Dunning commented on MAHOUT-1059:
-------------------------------------

I have added a fix/enhancement to the basic matrix to make handling of size caching more generic.  The basic issue was that caching the squared length of a vector makes distance calculations with L_2 faster for sparse vectors.  This should apply (or not) to all view-like vectors like DelegatingVector.  So now it does.
                
> New matrix extensions
> ---------------------
>
>                 Key: MAHOUT-1059
>                 URL: https://issues.apache.org/jira/browse/MAHOUT-1059
>             Project: Mahout
>          Issue Type: Improvement
>          Components: Math
>            Reporter: Ted Dunning
>             Fix For: 0.8
>
>         Attachments: 0001-MAHOUT-1059-Added-Centroid-WeightedVector-Delegating.patch, 0001-MAHOUT-1059-Added-Centroid-WeightedVector-Delegating.patch, 0001-MAHOUT-1059-Added-Centroid-WeightedVector-Delegating.patch, 0001-MAHOUT-1059-Added-Centroid-WeightedVector-Delegating.patch, 0002-MAHOUT-1059-Stylistic-cleanups.patch, 0002-MAHOUT-1059-Stylistic-cleanups.patch, 0002-MAHOUT-1059-Stylistic-cleanups.patch, 0003-MAHOUT-1059-Add-generic-vector-test.patch, 0003-MAHOUT-1059-Add-generic-vector-test.patch, 0004-MAHOUT-1059-Indentation.patch, 0004-MAHOUT-1059-Indentation.patch, 0005-MAHOUT-1059-Abstract-the-idea-of-a-cached-length.patch, 0006-MAHOUT-1059-Additional-test-for-weighted-vectors.patch, DelegatingVectorTest.java
>
>
> The upcoming clustering needs several capabilities to support different operations.  These include some matrix extensions for adding behaviors to different kinds of matrices.  Also there is a file based matrix that uses mmap to access a file as if it were a matrix in shared memory.  Since this is off-heap and shared between processes, it can seriously help some programs.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira