Artificial Immune Computing Model Of Multi-group Competition And Collaboration And Its Application

Co-evolution algorithm is a hot research topic in the field of computational intelligence in recent years. It refers to co-evolution mechanism in nature. In the algorithm, it uses multiple sub-populations evolved alone to simulate several species in nature, and makes these sub-populations interacted and influenced each other in the process of evolution. Heuristic cooperative model has been successfully used in many fields, such as biology, physics, chemistry, economics, anthropology and psychology, and even large-scale NP problem. Compared with the traditional evolutionary algorithm, its advantages are obvious. Nevertheless, the co-evolution algorithm also has some shortcomings difficultly to overcome. For example, its mechanism and model are complex than traditional evolutionary algorithm, the characteristics of the problem itself is very dependent, can not guarantee the performance of algorithms. Because of these deficiencies, the co-evolution algorithm is difficult to achieve good results on some simple question, and its sensitivity to parameters is also restricting applications of the algorithms.To resolve the problem of co-evolution algorithm, the paper draws on relevant principles of immune computation, and proposes a co-evolution model fused with the immune mechanism (ICEA). The model divides antibodies into several populations, through the introduction of selection, mutation, crossover, migration, immune and other operators, making the composition synergies between the various groups. These populations evolve respectively, and at the same time interact with each other. The result of simulation experiment with 13 benchmark functions verifies the feasibility and effectiveness of the model.Then, the ICEA algorithm is extended. The paper introduces the competition mechanism of nature into the intelligent evolutionary algorithm, and proposes a Lotka-Volterra model with finite resource control. The environmental resources gross of the model is set to a fixed value and all species need compete for environmental resources. In the model, each population were given a rating based on the evaluation value, individuals in dominant population can be growth in the number, inferior species will be limited to its development. Therefore, the populations survive in a natural environment of the certain number of resource constraints, through competition and collaboration among each other and driving each other to improve the complexity and performance, to achieve the competitive co-evolution between species.Finally, for the problem of dimension reduction of high dimensional data before clustering and analysis, a co-evolution algorithm based on immune projection pursuit is put forward (ICEA-PPC). The model introduces principles of evolutionary algorithms to solve the problem of projection pursuit dimension reduction, and it takes advantage of immune co-evolution algorithm to optimize projecting direction, so that the high-dimensional data samples are projected to low dimensional space. Such operations realize the aim that it projects the data from high dimensional space to a low one, which not only reduces the computational complexity of data mining process, but also makes the data shrinking become possible. And the experimental results demonstrate the validity of the proposed algorithm.