Research On Reinforcement Learning Mechanism Based On Artificial Immune Network

Intelligent information retrieval is one of the most important applications in the times of internet. The existing theories and methods of machine learning are difficult to adapt to the dynamic of data and the variety of user interests in the network environment, and become a weak link of the research of intelligent information retrieval. Therefore, a highly adaptive learning mechanism which can continuously learn from the dynamic data is needed. This paper tries to provide a solution for the parameters adjustment in the intelligent information retrieval model, by using the new soft computing technology artificial immune networks.The human immune system is the important barrier to protect the body from the foreign pathogen, and has many characteristics, such as self-recognition, immune memory, continuously improving in the process of life and so on, which give the computer scientists much inspiration to solve the engineering problems. Benefiting from the mechanism and principles of human immune system, we can study and implement the artificial immune system to solve the practical engineering problems. This is a very valuable research. The theories and methods of the artificial immune system is particularly applicable to the processing of network information.By imitating from the characteristics and principles of the natural immune system, this paper studies and designs the mathematical model, architecture, triggering strategy and learning mechanism of the artificial immune network based on the danger theory, which is the newest theoritical hypothesis about the triggering mechanism of human immune system. We design and implement the artificial immune network algorithm, and successfully apply this algorithm in solving a pattern recognition problem and a data clustering problem. Finally, we apply this new machine learning mechanism to the reinforcement learning of Chinese lexical analysis, and establish the Chinese word segmentation model based on the artificial immune network and the Chinese part-of-speech tagging model based on the artificial immune network under a premise that the feedback information is available. Experiments show that this method achive good results and the artificial immune network is a promising machine learning mechanism.