基于ART-人工免疫网络的数据浓缩方法研究  

DATA ENRICHING BASED ON ART-ARTIFICIAL IMMUNE NETWORK

作　　者：杜海峰[1] 王孙安[1] 

机构地区：[1]西安交通大学机械工程学院,西安710049

出　　处：《模式识别与人工智能》2001年第4期401-405,共5页Pattern Recognition and Artificial Intelligence

基　　金：教育部青年骨干教师资助计划资助项目

摘　　要：基于智能互补融合观点,提出了一种新的数据浓缩方法。该方法整合了自适应共振网络(ART)和人工免疫网络各自的优点,人工免疫网络利用ART提供的有关系统的先验知识构造网络,对数据样本进行浓缩,获得原始数据集的内部映像,为进一步处理提供良好的数据样本。本文详细讨论了算法步骤,还提出新的数据浓缩度量指标——数据蒸发率。对R2空间分类和Fisher花瓣问题的实验结果表明,该方法有效克服了人工免疫网络计算量大等缺点,结果可靠,且数据蒸发率高。Based on the point view of complementary strategies, a new method of data enriching which integrates the advantages of Adaptive Resonance Theory and Artificial Immune Network is put forward. This method uses the transcendental information about the system to construct the Artificial Immune Network for enriching data. It obtains the internal images of the original data, and provides good samples for the next data managing step. The steps of the algorithm are discussed in this paper into details. Additionally, the criterion about the level of data enriching is put forward. Experiments on five class problem in R2 space and the Fisher Iris data indicate that the new method reduces the quantum of the Artificial Immune Network and achieves credible results and high data enriching rate.

关 键 词：自适应共振网络 人工免疫网络 数据浓缩 人工智能 ART 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]