jenshadlich / matrix_terminator Goto Github PK

This project use different evolutionary algorithms (ES, PSO, CCPSO2) and a local 
search implemented in CUDA to search for results for the termination competition
[1]. For every algorithm exists a stand alone program and they are part of a C 
library to use them from a high-level language like Haskell.   

The evolutionary strategy and the particle swarm optimization is working. The
CCPSO2 is based on the paper of Li and Yao [2]. The ccpso2 version is alpha 
quality, not well parallelized and not ready for prime time. The standard 
particle swarm optimization borrows the mutation of that paper.

The custom algorithm is the only one which finds a result for z001 alias
aabb > bbbaaa. 

All algorithms where tested with CUDA 3.2 and 4.1 on a GTX 295 with compute
capability 1.3 .

If you have any questions contact me by mail.

The implementation is licensed under GPL 2.

[1] http://www.lri.fr/~marche/termination-competition/
[2] http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=5910380&abstractAccess=no&userType=inst