Modeless user interface incorporating automatic updates for developing and using bayesian belief networks

• US 20060020568A1
• Filed: 07/18/2005
• Published: 01/26/2006
• Est. Priority Date: 07/26/2004
• Status: Active Grant

• Current Assignee: Charles River Analytics, Inc.
• Original Assignee: Charles River Analytics, Inc.
• Inventors: Cox, Zachary T., Pfautz, Jonathan D.
• Granted Patent: US 7,536,372 B2
• Time in Patent Office: Days
• US Class Current: 706/52
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

An application for developing and using a model of a Bayesian Network to compute beliefs. The application provides an interface through which a user may specify the construction of the Bayseian Network, such as by specifying nodes in the network, parameters associated with the nodes, conditional probability distributions associated with the parameters or evidence that a parameter has a particular value. The application builds an inference engine based on user input specifying the construction of the Bayesian Network and uses it to compute beliefs. The application provides a user interface through which a user may specify the construction of the Bayesian Network and automatically updates an output reflecting beliefs. The input and output information may be available to the user simultaneously without switching operating modes of the application.

Claims

1. A method of processing data for a belief network, comprising the acts of:

a) receiving a plurality of inputs representing evidence and information indicative of a structure of a belief network;

b) providing at least one output representative of at least one belief based on the plurality of inputs; and

c) automatically updating the at least one output representative of at least one belief in response to at least one of the plurality of inputs having a value representative of a change to the evidence or the information indicative of the structure of the belief network.

2. The method of claim 1, wherein the belief network is modeled as a plurality of interconnected nodes and receiving information indicative of a structure in the act a) comprises receiving information indicative of interconnections between the nodes and at least one conditional probability distribution associated with a node.

3. The method of claim 2, wherein the act c) of automatically updating comprises the act of automatically updating in response to an input representative of a change in the at least one conditional probability distribution.

4. The method of claim 2, wherein the act c) of automatically updating comprises the act of automatically updating in response to one of the plurality of inputs representing a change in the evidence.

5. The method of claim 1, additionally comprising, prior to the act c) of automatically updating the output, the act of:

d) selectively placing the belief network development application in an operating state in which the outputs are automatically updated in response to an input representative of a change to the evidence or the information indicative of the structure of the belief network.

6. The method of claim 1, wherein the act c) of automatically updating comprises the act of automatically updating the output representative of at least one belief in response to one of the plurality of inputs representing a change within only a predetermined subset of inputs.

7. The method of claim 6, wherein the act c) of automatically updating comprises automatically updating the output representative of at least one belief in response to one of the plurality of inputs representing a change in evidence or a change in a conditional probability distribution.

8. The method of claim 6, additionally comprising the act of:

d) receiving an input from a human user representative of the predetermined subset of inputs.

9. The method of claim 1, wherein the acts a), b), and c) are performed by a belief network development application operating on a computer to which a data generating application is coupled and the method additionally comprises the act of:

d) receiving one of the plurality of inputs having a value representative of a change from the data generating application.

10. A software product comprising a plurality of computer-executable modules, comprising:

a module for receiving information representative of a BN;

a module for detecting received information representative of a modification of the BN; and

a module for building an inference engine for computing beliefs based on the BN in response to the detected modification of the BN.

11. The software product of claim 10, wherein the BN is modeled as a plurality of interconnected nodes and the module for detecting modifications of a BN detects information that represents a change to the model of the interconnected nodes.

12. The software product of claim 10, wherein the BN is modeled as a plurality of interconnected nodes, each of which has a parameter associated therewith with information representing a conditional probability distribution of values of the parameter and the module for detecting modifications of a BN detects information that represents a change to the information representing the conditional probability distribution.

13. The software product of claim 10, wherein the BN is modeled as a plurality of interconnected nodes, each of which has a parameter associated therewith with information representing evidence that the parameter has a value and the module for detecting received information representative of a modification comprises a module for detecting received information representative of a change in the evidence of a parameter of at least one of the plurality of interconnected nodes.

14. The software product of claim 10, wherein the module for building an inference engine in response to a detected modification of the BN builds the inference engine only when the detected modification is within a portion of the BN.

15. The software product of claim 14, additionally comprising a module for receiving a user specification of the portion of the BN.

16. The software product of claim 10, additionally comprising a module providing a user interface through which a user may input a command to control the inference engine to compute beliefs based on the BN, the user interface available for display in any operating state of the software product.

17. The software product of claim 10, wherein the module for detecting received information representative of a modification of the BN is operable in any operating state of the software product.

18. The software product of claim 10 wherein the module for receiving information representative of a BN is also adapted to display information about the BN which can automatically update in response to receipt of information representative of a BN.

19. The software product of claim 10, additionally comprising a plurality of display modules, each display module presenting on a display information reflective of beliefs computed by the BN and having an interface for receiving information indicative of changes in beliefs computed by the BN, and wherein each display module is adapted to automatically update the information reflective of beliefs on the display in response to receipt of information indicative of a change in beliefs.

20. In a computer system having a graphical user interface including a display and a user input device, a method of displaying information to a user and receiving input from the user, comprising the acts of:

a) simultaneously;

i) displaying in a first viewing area information reflecting a construction of a belief network;

ii) displaying in a second viewing area information reflecting a belief derived from a computation using the belief network;

b) receiving user input based on an interaction with the first viewing area, the user input specifying a modification of the construction of the belief network;

c) automatically computing information reflecting a modified belief derived from a computation using the belief network as modified according to the user input; and

d) automatically displaying in the second viewing area information reflecting the modified belief.

21. The method of claim 20, further comprising the act of:

e) simultaneously with the act a), displaying in a third viewing area information reflecting evidence used in the act c) of automatically computing.

22. The method of claim 21, further comprising the act of:

e) simultaneously with the act a), displaying in a third viewing area information reflecting a construction of the BN, including nodes, interconnections among the nodes and conditional probability distributions for parameters associated with the nodes.

23. The method of claim 22, wherein the acts a), b), c), d) and e) implement a modeless user interface comprising user controls for providing the user input received in the act b) without switching modes.

24. The method of claim 21, wherein the first viewing area, the second viewing area and the third viewing area provide a modeless user interface for a belief network development application.

25. The method of claim 20, wherein each of the first viewing area and the second viewing areas has a control associated with it, the method further comprising the acts of:

e) receiving a control input from the user actuating the control associated with the first viewing area or the second viewing area; and

f) in response to receiving the control input from the user, hiding the viewing area associated with the actuated control.

26. The method of claim 20, wherein each of the first viewing area and the second viewing area has a control associated with it, and the method further comprises the acts of:

e) receiving a control input from the user actuating the control associated with the first viewing area or the second viewing area;

f) in response to receiving the control input from the user, hiding a portion of the information in the viewing area associated with the actuated control.

27. The method of claim 20, additionally comprising the acts of:

e) receiving a user input specifying a fixed location of the first viewing area and displaying the first viewing area in the fixed location; and

f) receiving user input specifying a floating location of the second viewing area and displaying the second viewing area in the floating location.

28. The method of claim 20, wherein the first viewing area comprises a slider control for receiving input concerning evidence that a parameter has a value and the method additionally comprises the act of:

e) receiving an input through the slider control and automatically displaying in the second viewing area a belief derived from the input received through the slider control.

29. The method of claim 28, wherein the acts a), b), c), d) and e) implement a BN development application and the belief derived from the input received through the slider control is displayed without switching operating modes of the BN development application.

30. The method of claim 20, wherein:

the belief network is modeled as a plurality of nodes, each having a parameter with a conditional probability distribution associated therewith; and

the act b) of receiving user input comprises receiving user input specifying a change in the type of a conditional probability distribution of a node.

31. The method of claim 20, wherein the act c) of automatically computing comprises automatically computing with an inference engine.

32. The method of claim 20, wherein the user interface further comprises a control and the method additionally comprises the act of:

e) computing information reflecting a belief in response to user acutation of the control.

33. The method of claim 20, wherein the act b) of receiving user input and the act d) of displaying occur in the same user interface mode.