贝叶斯网络推理算法的研究和实现

Research and implementation of inference algorithm in Bayesian networks

针对桶消元推理算法提出并行计算的改进方案,并构建网络实验平台以用于研究中的理论验证及应用开发。实验表明,改进后的算法确实可以提高原算法的计算速度。

This paper puts forward the improved scheme of parallel computing for the bucket-elimination(BE) algorithm,and builds an integrated experimental platform for the test of inference and implemention of a Bayesian network. Experimental results indicate the improved scheme is more effective than the traditional BE in computation speed.