環境変化に適応するためのピボット型一般化

抄録

This paper proposes a concept and mechanism of the pivot-based generalization which can adapt to changing environment by introducing s# which represents two opposite situations while # which represents one situation in the context of Learning Classifier Systems. Intensive experiments, have revealed that (1) the pivot-based generalization can generalize individuals in the 2-objective Knapsack problem and real world water bus route optimization problem; (2) the sharp distance which represents distance between each individuals of the generalize individual contributes to controlling the number of s# individuals; (3) the pivot-based generalization of individuals can clarify the feature of solution space.