Adaptive parallel implementation of the Combinatorial GMDH algorithm. O.A. Koshulko, A.I. Koshulko. IWIM, Prague, 2007.

Abstract. The combinatorial algorithm of the Group Method of Data Handling is a compute intensive modelling method well proven for analysis and forecast of a variety of complex systems especially of the so-called “black-box” type. The algorithm is highly dependent on computational power that makes the use of multiprocessing reasonable. To exploit efficiently different kinds of multiprocessor computer systems we propose an adaptive parallel implementation of the combinatorial GMDH algorithm and an example of its usage for the forecasting of the “Top500 Supercomputer’s List”.

Keywords. Combinatorial GMDH, parallel processing, Parallel COMBI, Top500 Supercomputer’s List, multiprocessor architectures.

References.

1. Madala H.R., Ivakhnenko A.G.: Inductive Learning Algorithms for Complex Systems Modeling. – CRC Press, 1994. – 368 p.

2. Ivakhnenko A.G., Koppa Y.V., Stepashko V.S.: Spravochnik po tipovim programmam modeluvannya. Kiev: Technika, 1980. – 184 p.

3. Dongarra J., Luszczek P., Petitet A.: The LINPACK Benchmark: Past, Present, and Future Concurrency and Computation: Practice and Experience – 2003. – Volume 15. – P. 1-18.