基于记忆原理的人际关系优化算法

Memory Principles-based Interpersonal Relationship Optimization

为求解复杂函数优化问题,基于人类记忆原理和人际关系,提出了一种新型函数优化方法,即MP-IRO算法。在该算法中,将个体分为恋人、知己、敌人、小人、陌生人5种对象类型,对应于恋爱、聚集、攻击、排斥、防御等5种行为,并构造相应的演化算子。恋爱算子能优先选择拥有长时记忆的个体,分享其表征特性;聚集算子能使个体摆脱局部最优解的陷阱;攻击算子能使个体之间活跃度增强;排斥算子能让个体远离瞬时记忆试探解方向,扩大搜索范围;防御算子能增加随机性。测试结果表明,本算法对求解复杂函数优化问题具有较高的适应性和收敛速度。

In order to solve the complicated function optimization problems, a new optimization algorithm was constructed based on the memory principles （MP） and interpersonal relationship （IR）, namely memory principles-based interpersonal relationship optimization （MP-IRO）. There are five object types in interpersonal relationship system, including lovers, friends, enemies, villains and strangers which successively correspond to loving, gathering, attacking, rejecting, defending instincts and five operators; the loving operator gives preference to individuals which memory type is long memory（LM） and shares CP with other individuals; the gathering operator could make individuals avert local optima; the attacking operator enables an individual＇s vitality to increase; the rejecting operator can be far from instantaneous memory to expand the search coverage; the defending operator increases the randomness. Results show that the algorithm has characteristics of strong search capability and high adaptability for the complicated function optimization problems.