Abstract A method is proposed to evaluate the uncertainty of multi-data of radiation effect experiments. Combining the radial basis function neural network with mixture model, a RBF-based mixture density network is constructed and a learning algorithm is given. With the proposed method, the relationship and the uncertainty of multi-data of radiation effect experiments can all be calculated. Examples show the effectiveness of the proposed method.
HAN Feng,WANG Jian-guo,QIAO Deng-jiang, et al. ANN Based Method for Evaluating the Uncertainty of Multi-data[J]. Acta Metrologica Sinica, 2011, 32(6): 564-569.
［1］范如玉,韩峰,郭红霞.基于QMU方法的电源系统抗γ总剂量辐射能力评估［J］.强激光与粒子束,2011,23(2):536~540.
［2］冯小娟.二元线性回归的不确定度评定方法［J］.西安建筑科技大学学报,2000,32(3):304~306.
［3］王惠文,孟洁.多元线性回归的预测建模方法［J］.北京航空航天大学学报,2007,33(4):500~504.
［4］焦李成.神经网络系统理论［M］.西安:西安电子科技大学出版社,1990,26~52.
［5］朱坚民,王中宇,吕延庆,等.基于神经网络的测量模型的建立与检验［J］.光学精密工程,2000,8(4):389~393.
［6］闫平凡,张长水.人工神经网络与模拟进化计算［M］.北京:清华大学出版社,2000,282~287.
［7］Bishop C M. Mixture Density Networks. Neural Computing Research Group Report［EB］.http://www.ncrg.aston.ac.uk,1994.
［8］袁亚湘,孙文瑜.最优化理论与方法［M］.北京:科学出版社,1997,219-240.
2011, Vol. 32 
