复数免疫计算的自适应与协同

项目名称： 复数免疫计算的自适应与协同

项目编号： No.61203325

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 自动化学科

项目作者： 高尚策

作者单位： 东华大学

项目金额： 25万元

中文摘要： 免疫计算是受生物免疫系统启发解决实际问题的智能方法，对解决工程领域中复杂实际问题具有重要的理论意义和应用价值。为了提高免疫计算在优化和模式识别等问题上的应用性能，本项目围绕免疫计算模型的复数编码形式，开展用于提高复数免疫模型的计算效率和鲁棒性能的自适应理论和协同机制的研究。通过继续深度挖掘生物免疫系统的隐喻思想以及对现有模型的理论分析，分别建立基于内积学习法则的全覆盖复数免疫网络模型，基于实数矩阵表达的复数免疫迭代型计算模型，以及混合复数免疫算法的协同计算模型，并搭建统一的复数免疫计算实验平台；进一步设计出基于群体性能指标的参数自适应调整方法，基于渐进衰退高斯函数的多样性控制策略，以及融合分布式估计算法的协同计算理论，从而提高复数免疫模型的计算效率，扩大其在不同应用场景下的计算鲁棒性能，为解决复杂工程问题提供新的研究思路、理论基础及应用手段。

中文关键词： 智能计算；免疫算法；群体智能；自适应机制；协同计算

英文摘要： Immune computing is one of the computational intelligent methods, which is inspired by the natural immune principles to solve real-world problems. It provides significant effects of baisic theories and applications on solving complex engineering problems. For the purpose of improving the performance of immune computing when applied to optimization as well as pattern recognition problems, this project concentrates on the research of the complex-valued encoding in immune computing, and thereafter the design of self-adapted control together with co-evolutionary method aiming to improve the effectiveness and robustness of the constructed complex-valued immune models. First, this project establishes a uniform complex-valued immunological experiment platform, wherein a complete complex-valued immune network based on the inner product learning rules, an iterated complex-valued computing model based real-valued matrix representation, and a hybrid co-evolutionary immune model are embedded. Then, in order to further improve the performance of the proposed models, a population-property-index related parameter self-adapted control, a population-diversity manipulation strategy based on the regression Gaussian function, and a co-evolutionary theory with the estimation distribution algorithm are designed. Finally, the construc

英文关键词： Intelligence Computing；Immune Algorithm；Swarm Intelligence；Self-adaptive Mechanism；Co-evolutionary Computation