A New Learning Algorithm Based on SGA Bayesian Network
Tiejun Jia, Qiang Sun · 2008
In this paper, the developed approach to Bayesian network construction based on Self-organizing Genetic Algorithm (SGA) from knowledge base is proposed to solve the problem that a typical characteristic of Bayesian network topology is dependences of each variable within the network and makes it impossible to optimize variables. In order to avoid an early convergence for a normal GA algorithm, the self-organizing organism is introduced and an effective operator is provided to search the global optimum value. At last the experiment results and the convergence of SGA are discussed.