证据的信任函数表示

证据的信任函数表示 ON REPRESENTATION OF EVIDENCE BY BELIEF FUNCTIONS 杜劲松 1 first-author 张尧庭 1 武汉大学管理学院 武汉大学管理学院 College of Business Administration, Wuhan University College of Business Administration, Wuhan University 以诊断推理为应用背景，研究了信任函数理论框架下的证据表示．从似然原理出发，给出了基于似然性的信任函数所要满足的性质，引入了嵌套表示，建立了嵌套表示与稳健贝叶斯推断的联系，并以一个简化的诊断推理过程为例，说明了嵌套表示的应用，最后指出嵌套表示与其它表示方法的差异来自对似然原理的不同理解． The issue of representation of evidence under the framework of belief function theory is stuied in the context of diagnostic reasoning in this paper. Properties are given which should be satisfied by likelihood based belief functions according to Likelihood Principle. Embedding representation is introduced. The relationship between embedding representation and robust Bayesian inference is established. Applicatons of embedding representation are demonstrated through a simplified diagnostic reasoning process. Finally, it is indicated that the difference among varies of representations arises from different understanding to Likelihood Principle. 人工智能 信任函数 证据表示 似然性 artificial intelligence belief function representation of evidence likelihood TP18 国家攀登计划资助 1994-04-01 2021-04-01 4