Abstract: In order to improve Clonal Selection Algorithm (CSA) and make it theoretically mature, this paper adopted two random convergence measures: complete convergence and mean convergence to do the convergence analysis for the proposed algorithm named improved clonal Selection Algorithm for Multi-class Classification (Multi_CSA). It demonstrated that the Multi_CAS satisfied the sufficient condition for convergence to a global optimal solution. An experiment was also performed to validate the result.The paper proves that Multi_CAS meets the sufficient condition for convergence.The experiment shows that the algorithm will converge after several generations.It is concluded that Multi_CSA can converge within limited generation and it is a relatively mature algorithm.

Key words: artificial immune, clonal selection, classification, convergence

摘要： 为了完善克隆选择算法(CSA),使算法理论上成熟,利用两个随机收敛性度量:完全收敛和均值收敛, 证明基于多类数据分类的改进克隆选择算法(Multi_CSA)满足收敛到全局最优解的充分条件,并以实验数据进行验证。从理论上证明了Multi_CSA满足收敛的充分条件,实验方面也表明该算法在经过一定的代数后会收敛。理论和实验上均表明:Multi_CSA是一个能在有限代内收敛的较为成熟算法。

关键词: 人工免疫, 克隆选择, 分类, 收敛性