Neural network surrogate models of interactive genetic algorithms with individual'sinterval fitness

作者：Gong Dun Wei; Ren Jie^*; Sun Xiao Yan

来源：Control and Decision, 2009, 24(10): 1522-1525+1530.

• Evolutionary design
• Interaction
• Interactive genetic algorithm
• Lower limits
• Neural network surrogate models
• Surrogate model
• Training data
• Upper limits

This paper presents a neural network surrogate model of interactive genetic algorithms with an individual's interval fitness in order to solve the problem of user fatigue. The genotype and the fitness of individuals evaluated by the user are sampled to train a neural network to approximate the upper limit and the lower limit of an individual's interval fitness. The trained surrogate model is applied to evaluate individuals in the subsequent evolutions. The training data and the surrogate model are continuously updated during the evolutions to guarantee the performance in precision of the surrogate model. In addition, the performance of the proposed algorithm is quantitatively analyzed, which is applied to a fashion evolutionary design system. The results show that the proposed algorithm has more opportunities to look for satisfactory solutions on the condition of alleviating user fatigue.

• 出版日期2009
• 单位中国矿业大学（北京）