决策层特征融合decisionlevelidentityfusion.pptVIP

有必要就介绍书中IFFN的例子。 例举书上221页例子 例举书上掷骰子的例子。 举例解释焦元的含义，若子集A，且，，则称A为mass函数的焦元。m(A)指对焦元A的信度分配值，表示方案集A获得的证据支持程度。 DS具有处理不完备信息的能力，在军事中有较大应用。 有重叠的部分也要计算 Dempster-Shafer’s method Dempster’s rule of combination for multi sources is 第三十一页，共四十五页。 Dempster-Shafer’s method Dillard describes an algorithm for combing probability masses in a more complex situation. Dempster’s rules of combination are both commutative and associative. Hence data from sensors may be combined in a hierarchical manner. As a result a variety of parallel implementation might be developed. Dillard, R. A., “Tactical Inferencing with the Dempster-Shafer Theory of Evidence” The Asilomar Conference of Circuits, Systems, and Computers,1983, Naval Post Graduate School, Santa Clara, CA, pp. 321-316 For 2 or 3 sensors in a nonparallel implementation, the D-S technique requires approximately twice the computational effort of Bayesian inference. 第三十二页，共四十五页。 Generalized Evidence Processing (GEP) Theory Criticisms of D-S inferene Lack of rigor in defining evidence through independent observation Several issues frequently cited about D-S method. Thomopoulos proposed another generalization of Bayesian theory termed the generalized evidence processing (GEP) approach. 第三十三页，共四十五页。 Generalized Evidence Processing (GEP) Theory An example GEP assigns probability masses and combines those mass data based on a priori conditional probability of hypotheses H0 and H1. d0=hypothesis H0 is true d1=hypothesis H1 is true d2=hypothesis H0 or H1 is true 第三十四页，共四十五页。 Generalized Evidence Processing (GEP) Theory In D-S, the evidence is combined in accordance with the intersection of propositions In GEP, the combination is based on quantified impact of the resulting decisions. 第三十五页，共四十五页。 Heuristic methods for identity fusion Treat the identity fusion problem as if a group of humans were faced with a decision problem. Each human performs the task of a sensor. (group decision-making) Voting methods Scoring models Ordinal ranking