Research Of The Principles And Applications Of Artificial Immune Network

Artificial immune network is a bio-inspired computational model that uses ideas and concepts from the immune system theory to solve engineering and scientific problems. It has been an important research field of the principles and applications of artificial immune system. This dissertation focuses on artificial immune network and its mechanism of learning and memory for data clustering and optimization environment. Models and algorithms proposed in this paper are application-oriented, that is to improve current artificial immune networks in a certain application field or to exert an artificial immune network to a new field. The main contributions of this dissertation can be concluded as follows.1. Fuzzy clustering is very useful as one of the steps in exploratory data analysis. The classical FCM algorithm has been widely used in many areas such as image segmentation, pattern recognition and gene expression data analysis. However, the method is sensitive to the selection of initial centers and tends to stop at local optima. A fuzzy clustering method based on artificial immune network called AINFCM is proposed to overcome the shortage of FCM. Through evolution of memory cells in the artificial immune network, clustering centers and membership matrix are updated, and an optimal solution of fuzzy clustering is outputted. Experiments with bench datasets show that the AINFCM artificial immune network outperforms FCM as well as artificial immune evolution algorithm for fuzzy data clustering according to several validity measures.2. Human immune system is very complex and intrinsically parallel. Recently, there have been very few works in the parallelization of immune-inspired algorithms. Therefore, it is necessary to study paralleling artificial immune system. Time complexity of the AINFCM algorithm indicates that fitness calculation of antibodies is time costing. Subsequently, a paralleling model named PAINFCM of the AINFCM artificial immune network is designed, which acts as master-slave pattern. That is to say, the artificial immune network is implemented in the master processor, while fitness calculation of antibodies is divided into many slave processors. Tests have proven that parallelization is an effective way to improve efficiency and to maintain diversity of artificial immune network.3. Recent researches have revealed that absence of stimulation and cooperation within network cells will affect searching ability and searching speed of artificial immune networks. In this paper, a cooperative artificial immune network model hybrid with cooperative mechanism of swarm intelligence is proposed. A cooperative artificial immune network with particle swarm behavior denoted as CoAIN is implemented for multimodal optimization. Memory cells of the CoAIN artificial network interact not only by competition but also by cooperation with sharing experience. Numeric benchmark functions were used to assess the performance of the CoAIN. Compared with opt-aiNet, BCA, hybrid GA, and PSO algorithms, the CoAIN artificial immune network present good performance in terms of optimal searching ability and running speed.4. A new method for parameter optimization of pharmacokinetics based on artificial immune network named PKAIN is proposed. To improve local searching ability of the artificial immune network, a partition-based concurrent simplex mutation is developed. By means of evolution of network cells in the PKAIN artificial immune network, an optimal set of parameters of a given pharmacokinetic model is obtained. In this paper, the Laplace transform was applied to the pharmacokinetic differential equations of remifentanil and its major metabolite, remifentanil acid. The PKAIN method was used to optimize parameters of the derived compartment models. Compared with residual and Gauss-Newton methods, the PKAIN algorithm provides a new solution to parameter optimization of pharmacokinetics.