[jira] [Commented] (SPARK-2885) All-pairs similarity via DIMSUM

["Clive Cox (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SPARK-2885?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14118269#comment-14118269 ] 

Clive Cox commented on SPARK-2885:
----------------------------------

Hi.

 I am testing this and am having problems with SparseVectors and the call to RowMatrix.columnMagnitudes(); Its throwing Assertions when the row matrix is built from SparseVectors. Seems to work ok with DenseVectors. It does not do this everytime and I have not tracked down yet what is the particular error. Its happening in the Breeze code. 
 Has anyone else seen these issue?
Creating random small sets of SparseVectors I sometimes get exception like::
java.lang.AssertionError: assertion failed: 2 3 3 2 2 SparseVector((72,1.0), (80,1.0), (87,1.0), (2,1.0)) SparseVector((72,1.0), (80,1.0), (87,1.0), (2,1.0))
	at scala.Predef$.assert(Predef.scala:179)
	at breeze.linalg.operators.SparseVectorOps$$anon$164.apply(SparseVectorOps.scala:426)
	at breeze.linalg.operators.SparseVectorOps$$anon$164.apply(SparseVectorOps.scala:388)
	at breeze.linalg.operators.BinaryRegistry$class.apply(BinaryOp.scala:60)
	at breeze.linalg.VectorOps$$anon$14.apply(Vector.scala:250)
	at breeze.linalg.NumericOps$class.$colon$times(NumericOps.scala:44)
	at breeze.linalg.SparseVector.$colon$times(SparseVector.scala:48)
	at org.apache.spark.mllib.linalg.distributed.RowMatrix$$anonfun$columnMagnitudes$2.apply(RowMatrix.scala:405)
	at org.apache.spark.mllib.linalg.distributed.RowMatrix$$anonfun$columnMagnitudes$2.apply(RowMatrix.scala:403)


> All-pairs similarity via DIMSUM
> -------------------------------
>
>                 Key: SPARK-2885
>                 URL: https://issues.apache.org/jira/browse/SPARK-2885
>             Project: Spark
>          Issue Type: New Feature
>            Reporter: Reza Zadeh
>            Assignee: Reza Zadeh
>
> Build all-pairs similarity algorithm via DIMSUM. 
> Given a dataset of sparse vector data, the all-pairs similarity problem is to find all similar vector pairs according to a similarity function such as cosine similarity, and a given similarity score threshold. Sometimes, this problem is called a “similarity join”.
> The brute force approach of considering all pairs quickly breaks, since it scales quadratically. For example, for a million vectors, it is not feasible to check all roughly trillion pairs to see if they are above the similarity threshold. Having said that, there exist clever sampling techniques to focus the computational effort on those pairs that are above the similarity threshold, which makes the problem feasible.
> DIMSUM has a single parameter (called gamma) to tradeoff computation time vs accuracy. Setting gamma from 1 to the largest magnitude allows tradeoff of computation vs accuracy from low computation to high accuracy. For a very large gamma, all cosine similarities are computed exactly with no sampling.
> Current PR:
> https://github.com/apache/spark/pull/1778
> Justification for adding to MLlib:
> - All-pairs similarity is missing from MLlib and has been requested several times, e.g. http://bit.ly/XAFGs8 and separately by Jeremy Freeman (see https://github.com/apache/spark/pull/1778#issuecomment-51300825)
> - Algorithm is used in large-scale production at Twitter. e.g. see https://blog.twitter.com/2012/dimension-independent-similarity-computation-disco . Twitter also open-sourced their version in scalding: https://github.com/twitter/scalding/pull/833
> - When used with the gamma parameter set high, this algorithm becomes the normalized gramian matrix, which is useful in RowMatrix alongside the computeGramianMatrix method already in RowMatrix
> More details about usage at Twitter: https://blog.twitter.com/2014/all-pairs-similarity-via-dimsum
> For correctness proof, see Theorem 4.3 in http://stanford.edu/~rezab/papers/dimsum.pdf



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org