sandialabs / diamondsampling

Maximum All-Pairs Dot-product Search (MADS)

Grey Ballard, Ali Pinar, Tamara G. Kolda, and C. Seshadri

This software contains Matlab and C implementations for computing only the largest entries of a matrix-matrix product.

To cite this work, please use:

G. Ballard, T. G. Kolda, A. Pinar and C. Seshadhri, Diamond Sampling for Approximate Maximum All-pairs Dot-product (MAD) Search, 15th IEEE International Conference on Data Mining, ICDM '15, IEEE Computer Society, pp. 11-20, November 2015, http://dx.doi.org/10.1109/ICDM.2015.46.

License

Copyright 2016 Sandia Corporation. Under the terms of Contract DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains certain rights in this software.

This software is released under the BSD 2-Clause license. Please see the LICENSE file.

How do I run the code?

There are two directories:

• pure_matlab contains Matlab only code
• mex_matlab contains C/C++ code and requires mex'ing to compile and run, but it is more efficient
• Each directory has a README.txt file to help with setup and running

CSparse Library

The C/C++ code makes use of the CSparse Library data structures and subroutines. The full CSparse library is included with this distribution; see mex_matlab/CSparse/Doc/License.txt for License information.

Questions?

• Grey: [email protected]
• Tammy: [email protected]