Koshul'ko O. Technologies of parallel processing in the Group Method of Data Handling

The dissertation is devoted to investigation of parallel processing in The Group Method of Data Handling (GMDH) and implementation of effective parallelization in the combinatorial algorithm of GMDH. A universal parallel software of the combinatorial algorithm capable of working with various multiprocessor computers is designed.A new method of construction of the Flat Neighborhood Networks for cost-effective GMDH-oriented computer is proposed. An example of the GMDH-oriented system based on FNN and combinatorial algorithm software is built and its use demonstrated though the salvation of practical tasks. Key words: GMDH, FNN, multiparametric modeling, cost-effective compute cluster, parallel processing