基于最小熵产生选择策略的遗传算法研究

针对遗传算法(Genetic Algorithm,GA)容易陷入局部最优的问题,借鉴热力学非平衡定态下的最小熵增原理,提出一种基于最小熵增原理的热力学选择策略,使个体的选择不再完全依赖于适应值。通过最小熵产生选择策略使种群在保证收敛速度的同时保持多样性,有效避免了种群陷入局部最优。通过定义个体密度来度量种群多样性,利用精英策略驱动种群熵产生快速下降;当种群多样性过低时,使用基于最小熵产生的选择策略产生新种群以保证种群多样性。在0/1背包问题和数值测试问题上的实验结果均表明,该策略能很好地保证解集分布的均匀性,防止种群陷入局部最优。同时,该策略也可应用于目前较新改进的遗传算法中,对算法效率也有一定的改进,具有很好地普适性。

GENETIC ALGORITHM BASED ON MINIMUM ENTROPY PRODUCTION SELECTION STRATEGY

Genetic algorithm is easy to fall into local optimum.Based on minimum entropy generation principle in thermodynamic non-equilibrium steady state,we proposed a thermodynamic selection strategy,so that individual selection was no longer completely dependent on fitness.Through the selection strategy of minimum entropy generation,the population kept diversity while guaranteeing convergence speed,and effectively avoided the population falling into local optimum.Individual density was defined to measure population diversity and elite strategy was used to drive population entropy to decline rapidly.When the diversity of population was too low,the selection strategy was used to generate new population to ensure the diversity of population.The experimental results on 0/1 knapsack problem and numerical test problem show that the strategy can ensure the uniformity of solution set distribution and prevent the population from falling into local optimum.The strategy can also be applied to the current improved genetic algorithm,which also improves the efficiency of the algorithm and has good universality.