人工免疫系统中变异算法研究

Research on Mutation Algorithm in Artificial Immune System

杨进[1,2] 刘晓洁[1] 李涛[1]

在基于人工免疫理论入侵检测系统中，变异算法的缺点导致了较高的误报率和漏报率。该文提出了一种变异机制，能够加快进化速度，保存具有优势特征的物种，提高检测效率和准确性。该机制定义了自体／非自体的概念和形式化描述，给出了成熟细胞动态方程、亲和力累积方程。实验结果表明，当k取40附近值时，只要参数合适，TP值就能稳定在95％以上。该模型具有良好的实时性、自适应性和准确性，为构建新一代高效、合理的网络安全系统提供了一种有效方案。[著者文摘]

Aiming at the deficiencies of current mutation algorithm based on artificial immune system, such as fault positive is high and the efficiency is very low, an improved evolution optimization with mutation algorithm is proposed. This new algorithm introduces evolution operator into the mutation rule during the cells mutation process to improve the detection efficiency and to overcome the shortcomi.ng of the local optimum. And the concepts and formal definitions of self/non-self are given, mature-lymphocyte lifecycle, and affinity accumulation process are presented. Experimental results show that when the k set about 40 with the parameter adapted, the TP value can be stabilized on 95 %. And the algorithm greatly enhances the response rate and precision of detection and the proposed model has the features of real-time processing, self-adaptively, thus providing a promising solution for intrusion detection.[著者文摘]

基于小生境技术和聚类分析的人工免疫算法

Artificial Immune Algorithm Based on Niche Technology and Cluster Analysis

郝晓丽 谢克明

针对标准人工免疫算法存在的早熟收敛和后期收敛速度慢的问题，本文提出了一种基于小生境技术和聚类分析的改进的人工免疫算法。首先运用嵌入进化标记的小生境技术对初始种群进化，“排挤机制”有效地保持种群的多样性，防止了早熟，而标记种群的进化方向则加快了算法的收敛速度。其次聚类方法的应用使得在各极值点附近形成了聚类区域，在不同的聚类区域运用人工免疫的趋同算子和异化算子分别进行粗搜索和细搜索，以保证全局寻优的速度和精度。仿真结果表明，该改进算法较之标准免疫算法，有更快的收敛速度、更强的全局搜索能力和更好的寻优精度。[著者文摘]

Due to premature convergence and low speed of latter convergence in conventional artificial immune algo rithm, the new method is introduced in the paper which is improved artificial immune algorithm based on niche technique and cluster analysis. Firstly niche technique with revolutionary recording is taken to initial population. ＂exclusion mechanism＂ can maintain population diversity to avoid premature, and labeling evolution direction dynamically can improve convergence speed. Then cluster analysis is applied to obtain cluster areas nearby extremums. Different operators are taken in different areas respectively. Similar-taxis operator is employed to realize optimization within Cluster areas, while dissimilation operator between them. Parallel searching in coarser and finer layer can ensure the speed and precision of _global optimization. Simulation shows that the improved algorithm has higher convergence speed, better capability of-global searching and better optimization precision.[著者文摘]

用于约束优化的人工免疫响应进化策略

A Novel Evolutionary Strategy Based on Artificial Immune Response for Constrained Optimizations

公茂果[1] 焦李成[1] 杜海峰[1,2] 马文萍[1]

基于克隆选择学说及生物免疫响应过程的相关机理，探讨一种新的人工免疫系统模型——人工免疫响应，提出用于解决约束优化问题的人工免疫响应进化策略；基于算法网络拓扑结构的分析表明，新算法比传统的进化策略（μ，λ）-ES具有更大的收敛概率．对10个标准测试问题的测试结果表明，与采用随机排序的进化策略和采用动态惩罚函数的进化策略相比，新算法在收敛速度和求解精度上均具有一定的优势．[著者文摘]

Based on the clonal selection theory and mechanisms of biological immune response, a novel artificial immune systems model, Artificial Immune Response （AIR）, is discussed. And based on Artificial Immune Response a novel evolutionary strategy for constrained optimizations is put forward. Analysis of its network framework shows that the new algorithm is convergent with a higher probability than （μ,λ） evolutionary strategy. The experiments on 10 benchmark problems show that when compared with the （μ,λ） evolutionary strategies adopting stochastic ranking method and dynamic penalty function method, the new evolutionary strategy is capable of improving the search performance significantly no matter in convergent speed or precision.[著者文摘]

基于免疫记忆的人工免疫算法模型及其应用

王磊 肖人彬

本文在深入分析现有人工免疫算法模型优缺点的基础上，提出了一种基于免疫记忆机制的改进人工免疫算法模型ARTIA，该模型融合了由生物免疫系统启发而来的免疫记忆机制，包括联想记忆和迭代记忆两种，采用了多种策略以保护群体多样性，进而在数值试验的基础上对ARTIA算法模型的性能进行了分析和讨论，最后通过本质上可以归结为旅行商问题（TSP）的多目标组合优化工程实例－岩石钻孔机路径选择问题，验证了该算法的有效性，结论部分对全文作了总结并对今后研究工作进行了展望。

基于免疫学的入侵检测系统一般模型

陈云芳 王汝传

该文从人工免疫系统在入侵检测中的实际应用出发，首先归纳了人工免疲系统可供入侵检测借鉴的生物机理，接着给出了入侵检测问题的形式化描述，然后重点介绍了入侵检测系统中应用人工免疫原理的一般模型、检测步骤以及相关算法，最后展望了基于免疫学的入侵检测系统的发展趋势。[著者文摘]

人工免疫工程综述

谈英姿 沈炯 等

概述免疫系统基本概念，比较了三大生物信息处理系统（神经系统，遗传系统和免疫系统）的特性，在此基础上提出了人工免疫工程（artificial immune engineering,AIE)的概念，介绍了AIE的主要方法，包括基于免疫网络学说的人工免疫网络模型，基于免疫特异性的否定选择算法和基于接种疫苗及免疫多样性的免疫进化算法；阐述了AIE的应用现状，包括人工免疫网络在人工智能领域，免疫特异检测在计算机安全领域及免疫反馈和免疫鲁棒自适应特性在智能控制领域的多种典型应用，最后对AIE研究应用前景作了进一步的展望。

基于资源限制性人工免疫系统的多光谱遥感影像分类方法

Classification of Multi-Spectral Remote Sensing Image Based on Resource Limited Artificial Immune System

钟燕飞 张良培 龚健雅 李平湘

提出了一种基于资源限制性人工免疫系统（resouree limited artificial immune systems, RLAIS ）的多光谱遥感影像分类方法。该方法采用RLAIS对遥感影像分类中选取的感兴趣样区进行样本训练，得到全局聚类中心，利用聚类中心对遥感影像进行分类。实验证明，该方法在分类精度上优于传统方法，其总精度和Kappa系数分别达到了91%和0．88，具有实用价值。[著者文摘]

In this paper, some initial investigations are conducted to employ resource limited artificial immune system （RLAIS） for classification of multi spectral remote sensing image. The proposed method trains the samples of regions of interest using the RLAIS and obtains the optimal the centers of classification. Image classification task by RLAIS is attempted and the preliminary results are provided. By the experiment, it is demonstrated that the method of this paper is superior to the three traditional algorithms, and its overall accuracy and Kap pa coefficient reach 91% and 0.88 respectively.[著者文摘]

基于人工免疫系统的遥感图像分类

钟燕飞 张良培 龚健雅 李平湘

提出了一种基于人工免疫系统(Artificial Immune System)的分类方法。该方法首先应用人工免疫系统的克隆选择算法对样本进行自学习，得到全局最优的聚类中心，然后利用学习得到的聚类中心对整幅影像进行分类。由于人工免疫系统继承了生物免疫系统的自组织、自学习、自识别、自记忆的能力，从而使得人工免疫系统具有非线性的分类能力，并能够快速准确地得到全局最优解，克服了传统分类方法约束条件多，容易陷入局部最优的缺点。实验结果证明，该算法分类精度上优于传统的分类方法，总精度和Kappa系数分别达到了89．80％和0．8725，因而具有实用价值。