Discrete Differential Evolution の提案

概要

論文の詳細を見る
The Differential Evolution (DE), which is the population-based optimization technique, is one of the global optimizers for the continuous design variables such as the Particle Swarm Optimization (PSO). In the traditional DE, due to the mutation in the algorithm, the rounding-off and truncating are often employed to handle the discrete and integer variables. In this paper, the Discrete Differential Evolution (DDE) to handle the discrete or integer design variables is proposed. In the proposed DDE, the mutation is considered as the exchange possibility between two particles. By considering the mutation as the exchange possibility, it is easy and possible to handle the discrete and integer variables. In addition, the initialization of the population are also introduced in the proposed DE. It is possible to escape from local minimum by introducing the initialization of the population. The algorithm of the proposed DDE is very simple, and can be easily extend to the Mixed-Discrete Nonlinear Problems (MDNLPs). The proposed DDE can be applied to a variety of discrete and integer optimization problems. The validity is examined through typical benchmark problems.