Implementation of the algorithm of hierarchical cluster analysis on GPU by means of CUDA technology

In the paper, the algorithm of the hierarchical cluster analysis is considered and the method is proposed to transfer this algorithm onto the parallel multiprocessor system used on modern graphics processing units (GPUs). Within the frameworks of some natural assumptions, we have estimated the run time of the algorithm in a sequential case, in a parallel case for some abstract parallel machine and for GPU. The algorithm is implemented on CUDA, allowing us to carry out the hierarchical cluster analysis much faster, than on CPU.