Research And Application On Multi-Granularity Immune Network

This thesis is supported by the National Natural Science Foundation of China (grant 60273087) and Natural Science Foundation of Beijing (grant 4032009).Natural immune system, whose intelligence has been gradually known, is a kind of parallel, distributed and adaptive information process system with very high intellectual behavior. In the last few years, imitating its mechanism to design the artificial immune system has become the focus of the artificial intelligence research field. Now, some application achievements have been made in the field of abnormal pattern-recognition, such as fault diagnose, pattern classification, image recognition, network security, etc.In the artificial immune systems that have already been put forward at present, the pattern numbers that every detector can recognize are the same, that is to say, granularity is the same. The set of that detectors is named single-granularity detector set, which is unable to satisfy the requirement of the efficiency and precision at the same time. Utilizing detectors with different granularity to form multi-granularity detector set, we can expand the abnormal mode type that can be measured, improve the efficiency, and reduce false negative rate.This paper carries on research based on the thought above, and have made the following innovative achievements:1) Multi-granularity detector generating algirithm—MGDGA is proposed. Having studied the size of minimum effective single-granularity detector set andthe calculating method of its relevant parameters and analysized experiment data, this paper found that single-granularity detector set can not improve the precision and efficiency at the same time. Therefore, multi-granularity detector generating algorithm is proposed to produce multi-granularity detector set, which improves the precision of detection while guaranteeing the efficiency.2) Based on multi-granularity detector set, an immune agent—IAg is designed.Encapsulating multi-granularity detector by agent can realize the cooperation between detectors in set and reduce the complexity of the whole system. Based on MGDGA and the thought mentioned above, immune agent has been designed and...