Algorithm Of Bayesian Network Structural Learning Based On Quantum Genetic

Uncertainty reasoning has been a research hotspot in the field of Artificial Intelligence in recent years. As a method of describing uncertainty knowledge and reasoning, Bayesian Networks (BNs) are applied extensively in many fields. BNs has a solid statistics grounding, explicit semantic structure, flexible reasoning ability, convenient decision making mechanism, and efficient learning mechanism. Bayesian Network learning mainly includes structure learning and parameter learning. The network parameters can be got from data sets, the difficulty of which is lower. It is proved that the network structre learning is a NP hard problem. So, it is of important significance for construct BNs to research efficient algorithms of structure learning.The thesis deeply studied the structure learning of Bayesian Network, the main content of which is as follows:Firstly, a new method of Bayesian Network structural learning based on Quantum Genetic Algorithms (QGA) is proposed. QGA is a new optimization method that combines quantum computation with Genetic Algorithms (GA). Compared with the conventional algorithms, QGA has tended to solve NP-hard problems. It exploits the parallelism of quantum computation in order to speed up genetic procedures. In QGA the class fitness evaluation and selection procedures are replaced by operations of quantum gates. It is of characteristics of rapid convergence and good capability of global search. The chromosome is formed by quantum coding the structure of BN. It makes this chromosome as a complete independent solution space to evolve by taking quantum variance operation.Secondly, the chaos algorithm has enumeration, random and sensitivity for initial values. Except keeping the best individual of the last generation, all the other individuals are reproduced. Searching at the best point's neighborhood using chaos variables can increase the efficiency of learning.Thirdly, chromosome has been evolved by using quantum rotation gate. The magnitude of the angle parameter has an effect on the speed and the quality of convergence. So a fuzzy quantum genetic algorithm based on the fuzzy rules is proposed.Experiment results show that it is more efficient and precise to learn Bayesian Networks using QGA and improved QGA.