Adaptive Mechanisms for Parallelization of the Genetic Method of Neural Network Synthesis

When using multi-core systems, it is necessary to effectively use the mechanisms of parallelization the processes for the more productive using of computing power. Unlike other optimization technologies, genetic algorithms contains a population of trial solutions that are competitively managed using specific operators. Genetic algorithm is characterized by iterative training of a population of individuals. The power of genetic algorithms is enhanced using distributed computing. Such algorithms are called parallel genetic algorithms. The paper suggests a mechanism that allows increasing the productivity of the method by using the criteria selection and the multi-parent crossover. This will reduce memory consumption and speed up the work. Thus, the use of multicore systems will be more rational.