Klasifikace dokumentů s využitím GPU-PSO

Abstract

This thesis deals with the vector classification problem, focusing on text categorization using the relatively new stochastic evolutionary algorithm, Particle Swarm Optimization (PSO). This algorithm was used on the text classification problem in [4], where he achieves the best results in comparison with other algorithms mentioned there. The algorithm and the whole process is very computationally demanding. The primary goal of this thesis was to design and implement improvement in order to speed up whole process. The calculation speed up was possible because of parallel computing on GPUs. Implementation in CUDA uses the high computing performance of graphic cards in parallel processing tasks. Multiple acceleration of the process was achieved by executing critical parts on the GPU. The current state of the PSO classification and classification on the GPU is given. Performed experiments and results are evaluated.