Methods for monitoring conflicts in inference systems
First Claim
1. A method for determining valuation-based system characteristics of a physical system represented as a set of basic probability assignments (BPAs), the method comprising acts, performed on a data processing system, of:
- receiving a sequence of normalized basic probability assignments (BPAs) designated mi, for index i=1, . . . , n;
having intermediate combined, unnormalized BPAs designated m′
(n);
determining a valuation-based system characteristic selected from a group consisting ofa local cumulative normalization factor,a local incremental normalization factor,a local cumulative conflict measure,a local incremental conflict measure,a global cumulative normalization factor,a Type I global incremental normalization factor,a Type II global incremental normalization factor,a global cumulative conflict measure,a Type I global incremental conflict measure, anda Type II global incremental conflict measuredetermined from;
k12=k/(k1k2),wherek1 is a normalization factor for a first BPA m1,k2 is a normalization factor for a second BPA m2,k is an overall normalization factor, andk12 represents an incremental normalization factor;
outputting the valuation-based system characteristic;
whereby the valuation-based system characteristic provides information for evaluating the physical system.
1 Assignment
0 Petitions
Accused Products
Abstract
The present invention provides a system and a method to keep track of conflicts resulting from combining basic probability assignments (BPAs) without actually performing the normalization operation at every step of combination. This approach can be applied recursively, allowing the system to keep track of the normalization and conflict of the belief combinations inside a large belief inference system without ever performing a normalization operation as long as the computing system precision allows. The present invention can be applied to monitor conflicts within most inference systems, including the Valuation Based System (VBS) framework for Dempster-Shafer (D-S) evidential reasoning.
17 Citations
38 Claims
-
1. A method for determining valuation-based system characteristics of a physical system represented as a set of basic probability assignments (BPAs), the method comprising acts, performed on a data processing system, of:
-
receiving a sequence of normalized basic probability assignments (BPAs) designated mi, for index i=1, . . . , n; having intermediate combined, unnormalized BPAs designated m′
(n);determining a valuation-based system characteristic selected from a group consisting of a local cumulative normalization factor, a local incremental normalization factor, a local cumulative conflict measure, a local incremental conflict measure, a global cumulative normalization factor, a Type I global incremental normalization factor, a Type II global incremental normalization factor, a global cumulative conflict measure, a Type I global incremental conflict measure, and a Type II global incremental conflict measure determined from;
k12=k/(k1k2),where k1 is a normalization factor for a first BPA m1, k2 is a normalization factor for a second BPA m2, k is an overall normalization factor, and k12 represents an incremental normalization factor; outputting the valuation-based system characteristic;
whereby the valuation-based system characteristic provides information for evaluating the physical system.- View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19)
where; k(n) denotes the local cumulative normalization factor, cumulative to a current BPA mn; m′
(n) denotes an non-normalized, combined BPA, and is represented by a recursive relation m′
(n)=m′
(n-1){circle around (×
)}mn, having an initial condition m′
(0)=m0, where m0 is a normalized BPA initially residing in a node under consideration; andA denotes a subset of Θ
, where Θ
represents a collection of exclusive and collectively exhaustive outcomes of states of the physical system represented by a node under consideration for the valuation-based system.
-
-
3. A method as set forth in claim 2, wherein the valuation-based system characteristic is a local cumulative conflict measure, determined from:
-
C(n)=1−
k(n),where; C(n) denotes the local cumulative conflict measure, cumulative to a current BPA mn; and k(n) denotes the local cumulative normalization factor, cumulative to a current BPA mn.
-
-
4. A method as set forth in claim 2, where the valuation-based system is represented by a Markov Tree, and wherein the valuation-based system characteristic is a global cumulative normalization factor, determined from:
-
where kΣ
denotes the global cumulative normalization factor;m′
Σ
denotes an non-normalized, combined BPA, and is equal to m′
(n){circle around (×
)}mδ
, and mδ
represents a combination of messages from all neighboring nodes; andA denotes a subset of Θ
, where Θ
represents a collection of exclusive and collectively exhaustive outcomes of states of the physical system represented by a node under consideration for the valuation-based system.
-
-
5. A method as set forth in claim 1, where the valuation-based system characteristic determined is a local incremental normalization factor, determined from:
-
kn=k(n)/k(n-1),where; kn denotes the local incremental normalization factor; k(n) denotes the local cumulative normalization factor, cumulative to a current BPA mn; and k(n-1) denotes a local cumulative normalization factor, cumulative to a previous BPA mn-1.
-
-
6. A method as set forth in claim 5, wherein the valuation-based system characteristic is a local incremental conflict measure, determined from:
-
Cn=1−
kn;where; Cn denotes the local incremental conflict measure; and kn denotes the local incremental normalization factor.
-
-
7. A method as set forth in claim 1, wherein the valuation-based system characteristic is a local cumulative conflict measure, determined from:
-
C(n)=1−
k(n),where; C(n) denotes the local cumulative conflict measure, cumulative to a current BPA mn; and k(n) denotes the local cumulative normalization factor, cumulative to a current BPA mn.
-
-
8. A method as set forth in claim 1, wherein the valuation-based system characteristic is a local incremental conflict measure, determined from:
-
Cn=1−
kn;where; Cn denotes the local incremental conflict measure; and kn denotes the local incremental normalization factor.
-
-
9. A method as set forth in claim 1, where the valuation-based system is represented by a Markov Tree, and wherein the valuation-based system characteristic is a global cumulative normalization factor, determined from:
-
where kΣ
denotes the global cumulative normalization factor;m′
Σ
denotes an unnormalized, combined BPA, and is equal to m′
(n){circle around (×
)}mδ
, and mδ
represents the combination of messages from all neighboring nodes; andA denotes a subset of Θ
, where Θ
represents a collection of exclusive and collectively exhaustive outcomes of states of the physical system represented by a node under consideration for the valuation-based system.
-
-
10. A method as set forth in claim 9, where the valuation-based system characteristic is a Type I global incremental normalization factor, determined from:
-
kI,n=kΣ
/kΣ
-1;where kI,n denotes the Type I global incremental normalization factor; and kΣ
and kΣ
-1 are the global cumulative normalization factors for a current BPA (mn) and a previous BPA (mn-1).
-
-
11. A method as set forth in claim 9, where the valuation-based system characteristic is a Type II global incremental normalization factor, determined from:
-
kII,n=kΣ
/(k(n)kδ
);where kII,n denotes the Type II global incremental normalization factor; kΣ
denotes the global cumulative normalization factors for the current input BPA (mn);k(n) denotes the local cumulative normalization factor, cumulative to a current BPA mn; and kδ
denotes the cumulative normalization factor for mδ
, which is a combined BPA of all messages from neighboring nodes.
-
-
12. A method as set forth in claim 1, where the valuation-based system characteristic is a Type I global incremental normalization factor, determined from:
-
kI,n=kΣ
/kΣ
-1;where kI,n denotes the Type I global incremental normalization factor; and kΣ
and kΣ
-1 are the global cumulative normalization factors for the current input BPA (mn) and the previous input BPA (mn-1).
-
-
13. A method as set forth in claim 12, where the valuation-based system characteristic is a global cumulative conflict measure, determined from:
-
CΣ
=1−
kΣ
;where; CΣ
denotes the global cumulative conflict measure; andkΣ
denotes the global cumulative normalization factor.
-
-
14. A method as set forth in claim 12, where the valuation-based system characteristic is a Type I global incremental conflict measure, determined from:
-
CI,n=1−
kI,n;where CI,n denotes the Type I global incremental conflict measure; kI,n denotes the Type I global incremental normalization factor.
-
-
15. A method as set forth in claim 1, where the valuation-based system characteristic is a Type II global incremental normalization factor, determined from:
-
kII,n=kΣ
/(k(n)kδ
);where kII,n denotes the Type II global incremental normalization factor; kΣ
denotes the global cumulative normalization factors for the current input BPA (mn);k(n) denotes the local cumulative normalization factor, cumulative to a current BPA mn; and kδ
denotes the cumulative normalization factor for mδ
, which is the combined BPA of all messages from neighboring nodes.
-
-
16. A method as set forth in claim 15, where the valuation-based system characteristic is a Type II global incremental conflict measure, determined from:
-
CII,n=1−
kII,n;where CII,n denotes the Type II global incremental conflict measure; and kII,n denotes the Type II global incremental normalization factor.
-
-
17. A method as set forth in claim 1, where the valuation-based system characteristic is a global cumulative conflict measure, determined from the relationship:
-
CΣ
=1−
kΣ
;where; CΣ
denotes the global cumulative conflict measure; andkΣ
denotes the global cumulative normalization factor.
-
-
18. A method as set forth in claim 1, where the valuation-based system characteristic is a Type I global incremental conflict measure, determined from the relationship:
-
CI,n=1−
kI,n;where CI,n denotes the Type I global incremental conflict measure; and kI,n denotes the Type I global incremental normalization factor.
-
-
19. A method as set forth in claim 1, where the valuation-based system characteristic is a Type II global incremental conflict measure, determined from the relationship:
-
CII,n=1−
kII,n;where CII,n denotes the Type II global incremental conflict measure; and kII,n denotes the Type II global incremental normalization factor.
-
-
20. A computer program product for determining valuation-based system characteristics of a physical system represented as a set of basic probability assignments (BPAs) on a data processing system, the computer program product comprising a computer-readable medium having computer program instructions stored therein for causing a computer to perform operations of:
-
receiving a sequence of normalized basic probability assignments (BPAs) designated mi, for index i=1, . . . , n; having intermediate combined, unnormalized BPAs designated m′
(n);determining a valuation-based system characteristic selected from a group consisting of a local cumulative normalization factor, a local incremental normalization factor, a local cumulative conflict measure, a local incremental conflict measure, a global cumulative normalization factor, a Type I global incremental normalization factor, a Type II global incremental normalization factor, a global cumulative conflict measure, a Type I global incremental conflict measure, and a Type II global incremental conflict measure determined from;
k12=k/(k1k2),where k1 is a normalization factor for a first BPA m1, k2 is a normalization factor for a second BPA m2, k is an overall normalization factor, and k12 represents an incremental normalization factor; outputting the valuation-based system characteristic;
whereby the valuation-based system characteristic provides information for evaluating the physical system.- View Dependent Claims (21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38)
where; k(n) denotes the local cumulative normalization factor, cumulative to a current BPA mn; m′
(n) denotes an non-normalized, combined BPA, and is represented by a recursive relation m′
(n)=m′
(n-1){circle around (×
)}mn, having an initial condition m′
(0)=m0, where m0 is a normalized BPA initially residing in a node under consideration; andA denotes a subset of Θ
, where Θ
represents a collection of exclusive and collectively exhaustive outcomes of states of the physical system represented by a node under consideration for the valuation-based system.
-
-
22. The computer program product as set forth in claim 21, wherein the valuation-based system characteristic is a local cumulative conflict measure, determined from:
-
C(n)=1−
k(n),where; C(n) denotes the local cumulative conflict measure, cumulative to a current BPA mn; and k(n) denotes the local cumulative normalization factor, cumulative to a current BPA mn.
-
-
23. The computer program product as set forth in claim 21, where the valuation-based system is represented by a Markov Tree, and wherein the valuation-based system characteristic is a global cumulative normalization factor, determined from:
-
where kΣ
denotes the global cumulative normalization factor;m′
Σ
denotes an non-normalized, combined BPA, and is equal to m′
(n){circle around (×
)}mδ
, and mδ
represents a combination of messages from all neighboring nodes; andA denotes a subset of Θ
, where Θ
represents a collection of exclusive and collectively exhaustive outcomes of states of the physical system represented by a node under consideration for the valuation-based system.
-
-
24. The computer program product as set forth in claim 20, where the valuation-based system characteristic determined is a local incremental normalization factor, determined from:
-
kn=k(n)/k(n-1),where; kn denotes the local incremental normalization factor; k(n) denotes the local cumulative normalization factor, cumulative to a current BPA mn; and k(n-1) denotes a local cumulative normalization factor, cumulative to a previous BPA mn-1.
-
-
25. The computer program product as set forth in claim 24, wherein the valuation-based system characteristic is a local incremental conflict measure, determined from:
-
Cn=1−
kn;where; Cn denotes the local incremental conflict measure; and kn denotes the local incremental normalization factor.
-
-
26. The computer program product as set forth in claim 20, wherein the valuation-based system characteristic is a local cumulative conflict measure, determined from:
-
C(n)=1−
k(n),where; C(n) denotes the local cumulative conflict measure, cumulative to a current BPA mn; and k(n) denotes the local cumulative normalization factor, cumulative to a current BPA mn.
-
-
27. The computer program product as set forth in claim 20, wherein the valuation-based system characteristic is a local incremental conflict measure, determined from:
-
Cn=1−
kn;where; Cn denotes the local incremental conflict measure; and kn denotes the local incremental normalization factor.
-
-
28. The computer program product as set forth in claim 20, where the valuation-based system is represented by a Markov Tree, and wherein the valuation-based system characteristic is a global cumulative normalization factor, determined from:
-
where kΣ
denotes the global cumulative normalization factor;m′
Σ
denotes an unnormalized, combined BPA, and is equal to m′
(n){circle around (×
)}mδ
, and mδ
represents the combination of messages from all neighboring nodes; andA denotes a subset of Θ
, where Θ
represents a collection of exclusive and collectively exhaustive outcomes of states of the physical system represented by a node under consideration for the valuation-based system.
-
-
29. The computer program product as set forth in claim 28, where the valuation-based system characteristic is a Type I global incremental normalization factor, determined from:
-
kI,n=kΣ
/kΣ
-1;where kI,n denotes the Type I global incremental normalization factor; and kΣ
and kΣ
-1 are the global cumulative normalization factors for a current BPA (mn) and a previous BPA (mn-1).
-
-
30. The computer program product as set forth in claim 28, where the valuation-based system characteristic is a Type II global incremental normalization factor, determined from:
-
kII,n=kΣ
/(k(n)kδ
);where kII,n denotes the Type II global incremental normalization factor; kΣ
denotes the global cumulative normalization factors for the current input BPA (mn);k(n) denotes the local cumulative normalization factor, cumulative to a current BPA mn; and kδ
denotes the cumulative normalization factor for mδ
, which is a combined BPA of all messages from neighboring nodes.
-
-
31. The computer program product as set forth in claim 20, where the valuation-based system characteristic is a Type I global incremental normalization factor, determined from:
-
kI,n=kΣ
/kΣ
-1;where kI,n denotes the Type I global incremental normalization factor; and kΣ
and kΣ
-1 are the global cumulative normalization factors for the current input BPA (mn) and the previous input BPA (mn-1).
-
-
32. The computer program product as set forth in claim 31, where the valuation-based system characteristic is a global cumulative conflict measure, determined from:
-
CΣ
=1−
kΣ
;where; CΣ
denotes the global cumulative conflict measure; andkΣ
denotes the global cumulative normalization factor.
-
-
33. The computer program product as set forth in claim 31, where the valuation-based system characteristic is a Type I global incremental conflict measure, determined from:
-
CI,n=1−
kI,n;where CI,n denotes the Type I global incremental conflict measure; kI,n denotes the Type I global incremental normalization factor.
-
-
34. The computer program product as set forth in claim 20, where the valuation-based system characteristic is a Type II global incremental normalization factor, determined from:
-
kII,n=kΣ
/(k(n)kδ
);where kII,n denotes the Type II global incremental normalization factor; kΣ
denotes the global cumulative normalization factors for the current input BPA (mn);k(n) denotes the local cumulative normalization factor, cumulative to a current BPA mn; and kδ
denotes the cumulative normalization factor for mδ
, which is the combined BPA of all messages from neighboring nodes.
-
-
35. The computer program product as set forth in claim 34, where the valuation-based system characteristic is a Type II global incremental conflict measure, determined from:
-
CII,n=1−
kII,n;where CII,n denotes the Type II global incremental conflict measure; and kII,n denotes the Type II global incremental normalization factor.
-
-
36. The computer program product as set forth in claim 20, where the valuation-based system characteristic is a global cumulative conflict measure, determined from the relationship:
-
CΣ
=1−
kΣ
;where; CΣ
denotes the global cumulative conflict measure; andkΣ
denotes the global cumulative normalization factor.
-
-
37. The computer program product as set forth in claim 20, where the valuation-based system characteristic is a Type I global incremental conflict measure, determined from the relationship:
-
CI,n=1−
kI,n;where CI,n denotes the Type I global incremental conflict measure; and kI,n denotes the Type I global incremental normalization factor.
-
-
38. The computer program product as set forth in claim 20, where the valuation-based system characteristic is a Type II global incremental conflict measure, determined from the relationship:
-
CII,n=1−
kII,n;where CII,n denotes the Type II global incremental conflict measure; and kII,n denotes the Type II global incremental normalization factor.
-
Specification