Abstract:
A parallel algorithm for solving the problem of constructing of nonlinear models (mathematical expressions, functions, algorithms, programs) based on given experimental data, a set of variables, basic functions and operations is proposed. The proposed algorithm of the multivariant evolutionary synthesis of nonlinear models has a linear representation of the chromosome, the modular operations in decoding the genotype to the phenotype for interpreting a chromosome as a sequence of instructions, the multivariant method for presenting a multiplicity of models (expressions) using a single chromosome. A comparison of the sequential version of the algorithm with a standard algorithm of genetic programming and the algorithm of the Cartesian Genetic Programming offers advantage of the algorithm proposed both in the time of obtaining a solution (by about an order of magnitude in most cases), and in the probability of finding a given function (model). In the experiments on the parallel supercomputer systems, estimates of the efficiency of the proposed parallel algorithm have been obtained showing linear acceleration and scalability.
Fulltext:
PDF file (716 kB)
