Download presentation
Presentation is loading. Please wait.
Published byXavier Vaughn Modified over 10 years ago
1
Geometry of Dempsters rule NAVLAB - Autonomous Navigation and Computer Vision Lab Department of Information Engineering University of Padova, Italy Fabio Cuzzolin FSKD02, Singapore, November 19 2002
2
2 1 The talk zintroducing the theory of evidence 2 zpresenting the geometric approach: the belief space 3 zanalyzing the local geometry of Dempsters rule 4 zperspectives of geometric approach
3
1 The theory of evidence
4
4 zgeneralize classical finite probabilities A Belief functions znormalization B2B2 B1B1 zfocal elements
5
5 Dempsters rule zare combined by means of Dempsters rule AiAi BjBj A i B j =A zintersection of focal elements
6
2 Geometry of belief functions
7
7 zit has the shape of a simplex Belief space zthe space of all the belief functions on a frame zeach subset A A-th coordinate s(A)
8
8 Global geometry of zDempsters rule and convex closure commute zconditional subspace: future of s zexample: binary frame ={x,y}
9
3 Local geometry of Dempsters rule
10
10 Convex form of zDempsters sum of convex combinations zdecomposition in terms of Bayes rule
11
11 Local geometry in S 2
12
12 Constant mass loci zset of belief functions with equal mass k assigned to a subset A zexpression as convex closure
13
13 zintersection of all the subspaces Foci of conditional subspaces zit is an affine subspace zgenerators: focal points
14
14 4 …conclusions za new approach to the theory of evidence: the belief space zgeometric behavior of Dempsters rule zapplications: approximation, decomposition, fuzzy measures
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.