Knowledge-Based Genetic Algorithms and its Application in Multi-Sensor Fusion

In this paper, rough set theory (RST) was introduced to discovery knowledge hidden in the evolution process of Genetic Algorithm. Firstly it was used to analyze correlation between individual variables and their fitness function. Secondly, eigenvector was defined to judge the characteristic of the problem. And then the knowledge discovered was used to select evolution subspace and to realize knowledge-based evolution. The result of weight-value optimization of the neural network in multi-sensor information fusion system shows that this method is able to effectively improve the study efficiency and study precision for neural networks.