基于Multi-memory机制的克隆选择算法及其在模式识别中的应用

人工免疫系统研究中大多借鉴克隆选择原理来构建免疫识别算法。描述了Castro提出的克隆选择算法CLONALG的整体框架,指出其在大规模模式识别问题中的不可收敛性。在CLONALG的基础上设计了Multi-memory机制，并以模式识别为应用背景提出了新的基于Multi-memory机制的克隆选择的免疫算法MCA，提出并深入分析了记忆抗体训练过程中的关键因素——变异概率的计算公式。实验表明，采用MCA的免疫系统具有更强的泛化能力、更高的抗原识别率以及更能适应大规模问题的优良特性。

Clonal selection algorithm based on multi-memory mechanism with applications to pattern recognition

Currently in the research of artificial immune system, the clonal selection principle is commonly used in designing immune recognition algorithms. This paper described the general framework of CLONALG which was a clonal selection algorithm proposed by Castro, and pointed out that the convergence could not be guaranteed when it was applied to large-scale pattern recognition problems. A multi-memory mechanism was designed, based on which a new immune algorithm referred to as MCA was proposed and applied to pattern recognition problems. A new formula was proposed to calculate the mutation probability, which was a key factor in training memory antibodies. The results of experiments show that both the generalization capability and the recognition accuracy of MCA are better than that of CLONALG, and the MCA can be effectively applied to large-scale problems.