Immune Clonal Strategy Algorithms And Their Application | Posted on:2006-11-02 | Degree:Doctor | Type:Dissertation | Country:China | Candidate:R C Liu | Full Text:PDF | GTID:1118360182460118 | Subject:Circuits and Systems | Abstract/Summary: | PDF Full Text Request | Artificial Immune System (AIS) is a new intelligent method simulating naturalimmune system, AIS aims at using ideas gleaned from immunology in order to developsystems capable of performing a wide range of tasks in various research areas such asnoise tolerance, self-learning, self-organization and memory, thus having enormouspotential supplying novel methods to solve complex problems. AIS has been used incybernation, data processing, optimization anomaly detection and so on. It is becominganother research hotspot in the artificial intelligent techniques after neural network,fuzzy system and evolutionary computation.Inspired by the immune system, some novel artificial immune system algorithmsbased on the clonal selection and immune memory mechanism are presented in thisthesis, these algorithms include immune clonal strategy algorithm, immune memorydynamic strategy algorithm and immune memory strategy algorithm. And using thetheories of Markov Chain, it is proved that algorithms are convergent and a simpleanalysis about convergent rate of the immune clonal strategy algorithm is presented.The Applications of these algorithms to some difficult tasks such as numericaloptimization problems and combinatorial optimization validate their potential of solvingcomplex problem. The main work can be summarized as follows:1. Biologically-motivated information processing systems can be classified into:Artificial Neural Network, Evolutionary Computation (EC) and Artificial ImmuneSystem. Among these, artificial network and EC have been widely applied tovarious fields, but there have relative few applications of AIS because of itscomplexity. With a further opening out of the immune system, artificial immunesystem will play more important role in many fields. A systemic expatiation on thecorrelative mechanisms of the natural immune system is presented and the history,research areas and development directions of the artificial immune system arediscussed, some important immune algorithms are introduced.2. Antibody Clonal Selection Theory is very important for the immunology. It hasattracted a great attention of the Artificial Intelligence researchers for it has a seriesof characteristics such as memory, learning and evolution. However, application ofthe antibody clonal mechanisms in AIS is relative few. Based on the clonal selectiontheory, a novel artificial immune system algorithm, Immunity Clonal Strategyalgorithm (ICSA), which includes Immunity Monoclonal Strategy Algorithm (IMSA)and Immunity Polyclonal Strategy Algorithm (IPSA), is put forward. It is easy tofind that the essential of the clonal operator is to produce a variation populationaround the parents according to their affinity. Thus, the searching area is enlarged.Ulteriorly, the clonal operator maps a problem in a low dimension space (ndimensions) to a high one (Nc dimensions), and projects the results to the originalspace afterwards clonal selection. Therefore the problem can be solved better. Usingthe theories of Markov Chain, it is proved that ICSA is convergent.3. In order to simulate the clonal selection and immune memory mechanism, a novelartificial immune system algorithm, Immune Memory Dynamic Clone Strategy(IMDCS), is put forward. The features of the proposed algorithm lie in: (1) Affinityis chosen as evaluation function, including evaluating Ab-Ag affinity and Ab-Abaffinity;(2) By promoting or restraining the production of antibodies, the algorithmcan adjust the clonal sizes of antibody population and memory unit adaptively,which improves the antibody diversity and makes the algorithm explore the affinitylandscape efficiently;(3) memory unit fastens the affinity maturation. The results ofthe academic analysis and the experiments indicate that IMDCS has highconvergence speed and better local searching ability, and supply a new method tosolve the complex problems such as multimodal function optimization and 0-1knapsack problem.4. Not only simulate the clone in process of immune response, also simulate immunememory mechanism and clonal deletion of B lymphocyte cell in immune response,another artificial immune system algorithm, Immune Memory Clonal Strategy(IMCS) is proposed in this paper. The IMCS adopts real-valued encoding and sets upa separate memory units. By simulating the mechanisms such as self-adjusting,memory learning and adaptation in the natural immune system, the algorithmrealizes the evolution of antibody population and the evolution of memory unit atthe same time, and by using clonal selection operator, the global optimalcomputation can be combined with the local searching. It is shown that IMCS hasthe strong abilities in having high convergence speed, enhancing the diversity of thepopulation and avoiding the premature convergence to some degree, with thecomputer simulations to some benchmark problems such as high-dimensionalfunction optimization and Traveling Salesman Problem.5. Followed the novel intelligent algorithms proposed, some experiments should bemade to illuminate its ability of solving practical problem, otherwise, acorresponding theoretic should be supplied to explain the efficiency of the proposedalgorithm, which include a proof of convergence and analysis of convergent rate. Bydint of some results in evolutionary algorithm and the Markov chain theory, asystemic academic analysis of the convergence of ICSA in chapter 4 is given,moreover, this paper obtained some bounds on the convergent rate of ICSA. Threebenchmark function optimization problems are be used to test the performance ofthe ICSA, the simulation results also confirm the theoretical results upwards. | Keywords/Search Tags: | Evolutionary Computation, Natural Immune System, Artificial Immune System, Clonal Selection, Immune Memory, Clonal Deletion, Markov Chain, Convergence, Convergent Rate, Multimodal Function Optimization, Traveling Salesman Problem, Knapsack Problem | PDF Full Text Request | Related items |
| |
|