Document Clustering Using Multi-Objective Genetic Algorithms with Parallel Programming Based on CUDA
Jung Song Lee, Soon Cheol Park, Jong Joo Lee, Han Heeh Ham · 2014
In this paper, we propose a method of enhancing Multi-Objective Genetic Algorithms (MOGAs) for document clustering with parallel programming. The document clustering using MOGAs shows better performance than other clustering algorithms. However, the overall computation time of the MOGAs is considerably long as the number of documents increases. To effectively avoid this problem, we implement the MOGAs with General-Purpose computing on Graphics Processing Units (GPGPU) to compute the document similarities for the clustering. Furthermore, we introduce two thread architectures (Term-Threads and Document-Threads) in the CUDA (Compute Unified Device Architecture) language. The experimental results show that the parallel MOGAs with CUDA are tremendously faster than the general MOGAs.