GPU-Accelerated MapReduce-Based Self-Organizing Maps
C++ C
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
data
LICENSE
Makefile
README
io.cpp
mrsom.cpp
mrsom.cu
mrsom.h

README

GPU-Accelerated MapReduce-Based Self-Organizing Maps

Please note that this project has been superseded by

https://github.com/peterwittek/somoclu

Example usage with two GPUs:

mpirun -np 2 ./mrsom data/rgbs.txt data/rgbs

The input is a plain-text dense matrix, where the rows correspond to the instances.

You will need CUDA, MPI, and MR-MPI to compile the code. MR-MPI is available at:
http://www.sandia.gov/~sjplimp/download.html

The code was derived from the MR-MPI SOM implementation by Seung-Jin Sul:
https://github.com/andreyto/mr-mpi-som

The GPU setting method for a distributed system is based on GPMR by Jeff Stuart:
http://code.google.com/p/gpmr/

If you use this code, please cite: 
Wittek, P., Darányi, S.: A GPU-Accelerated Algorithm for Self-Organizing Maps in a Distributed Environment. Proceedings of ESANN-12, 20th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning. Bruges, Belgium. April, 2012.