|Budget Amount *help
¥2,300,000 (Direct Cost : ¥2,300,000)
Fiscal Year 1999 : ¥500,000 (Direct Cost : ¥500,000)
Fiscal Year 1998 : ¥1,800,000 (Direct Cost : ¥1,800,000)
Genetic algorithms (GA) are useful methods for complicated optimization problems.
However, their main drawback is computationally expensive. Therefore, parallel processing is inevitable, but the researches on parallel genetic algorithms are not enough. This research aims to evaluate the effectiveness of parallel genetic algorithms and to propose a new distributed GA which shows a high efficiency in parallel processing.
First, some distributed GA models with divided sub-populations are compared and the performance of the models are investigated. The effect of the GA parameters are investigated, and a new approach is proposed. The following are the summaries of the search.
1) The distributed population model with divided sub-populations with migration is suitable for parallel processing and its performance for providing good solutions in very high. The distribution population model suppress the early convergence.
2) The optimum crossover and mutation rates for the distributed GA are different from those for single population GA, and a new distributed GA model with distributed environment is proposed to relive the difficulty in choosing optimum parameters.