Incremental data fusion and decision making system and associated method

• US 20060112039A1
• Filed: 09/19/2005
• Published: 05/25/2006
• Est. Priority Date: 09/17/2004
• Status: Active Grant

• Current Assignee: Johnson Controls Tyco IP Holdings LLP (Johnson Controls International plc)
• Original Assignee: Proximex Corporation (Apollo Global Management, Inc.)
• Inventors: Wang, Yuan-Fang
• Granted Patent: US 7,467,116 B2
• Time in Patent Office: Days
• US Class Current: 706/20
• CPC Class Codes: G06F 18/256 of results relating to different input data, e.g. multimodal recognition
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06Q 10/00 Administration; Management

Abstract

A computer implemented adaptive ensemble classifier is provided which includes: a plurality of classifiers; a decision structure that maps respective classifier combinations to respective classification decision results; and a plurality of respective sets of weights associated with respective classifier combinations.

Claims

1. A computer implemented adaptive classification method comprising:

producing at least one first classification result using one or more first classifiers and one or more first data sources;

producing a first decision using the at least one first classification result and a first set of one or more respective weights respectively associated with the one or more first classifiers;

selecting at least one second classifier if the first decision has a first prescribed value;

producing at least one second classification result using the one or more first classifiers and the at least one second classifier and the one or more first data sources and at least one second data source; and

producing a second decision using the at least one second classification result and a second set of one or more respective weights respectively associated with the one or more first classifiers and with the at least one second classifier.

2. The method of claim 1 further including:

terminating the classification process if the first decision has a second prescribed value.

3. The method of claim 1 further including:

selecting at least one third classifier if the first decision has a second prescribed value;

producing at least one third classification result using the one or more first classifiers and the at least one third classifier and the one or more first data sources and at least one third data source; and

producing a third decision using the at least one second classification result and a third set of one or more respective weights respectively associated with the one or more first classifiers and with the at least one third classifier.

4. The method of claim 3 further including:

selecting at least one fourth classifier if the third decision has a first prescribed value;

producing at least one fourth classification result using the one or more first classifiers, the at least one third and the at least one fourth classifiers and the one or more first and third data sources and at least one fourth data source;

producing a fourth decision using the at least one fourth classification result and a fourth set of one or more respective weights respectively associated with the one or more first classifiers and with the at least one third and the at least one fourth classifiers.

5. The method of claim 3 further including:

terminating the classification process if the third decision has a second prescribed value.

6. A computer implemented adaptive ensemble classifier comprising:

a plurality of classifiers;

a decision structure that maps respective classifier combinations to respective classification decision results; and

a plurality of respective sets of weights associated with respective classifier combinations.

7. The adaptive ensemble classifier of claim 6,

wherein the decision structure maps respective classifier combinations to respective sequences of classification decision results.

8. The adaptive ensemble classifier of claim 6,

wherein the decision structure comprises a tree-structure.

9. The adaptive ensemble classifier of claim 6,

wherein the decision structure comprises a tree-structure;

and

wherein each respective classifier combination is associated with one or more branch of the tree structure.

10. The adaptive ensemble classifier of claim 6,

wherein the decision structure comprises a tree-structure;

and

wherein respective sets of weights are associated with respective branches of the tree structure.

11. A processing system comprising:

means for performing each of a plurality of classifications so as to produce classification decision results;

means for mapping respective classification combinations to respective classification decision results; and

means for storing a plurality of respective sets of weights associated with respective classifier combinations.

12. The processing system of claim 11,

wherein the means for mapping includes a tree-decision structure in computer memory.

13. A processing system comprising:

means for producing at least one first classification result using one or more first classifiers and one or more first data sources;

means for producing a first decision using the at least one first classification result and a first set of one or more respective weights respectively associated with the one or more first classifiers;

means for selecting at least one second classifier if the first decision has a first prescribed value;

means for producing at least one second classification result using the one or more first classifiers and the at least one second classifier and the one or more first data sources and at least one second data source; and

means for producing a second decision using the at least one second classification result and a second set of one or more respective weights respectively associated with the one or more first classifiers and with the at least one second classifier.

14. The system of claim 13 further including:

means for terminating the classification process if the first decision has a second prescribed value.

15. The system of claim 13 further including:

means for selecting at least one third classifier if the first decision has a second prescribed value;

means for producing at least one third classification result using the one or more first classifiers and the at least one third classifier and the one or more first data sources and at least one third data source; and

means for producing a third decision using the at least one second classification result and a third set of one or more respective weights respectively associated with the one or more first classifiers and with the at least one third classifier.

16. The system of claim 15 further including:

means for selecting at least one fourth classifier if the third decision has a first prescribed value;

means for producing at least one fourth classification result using the one or more first classifiers, the at least one third and the at least one fourth classifiers and the one or more first and third data sources and at least one fourth data source;

means for producing a fourth decision using the at least one fourth classification result and a fourth set of one or more respective weights respectively associated with the one or more first classifiers and with the at least one third and the at least one fourth classifiers.

17. The system of claim 15 further including:

means for terminating the classification process if the third decision has a second prescribed value.

18. A computer readable medium having computer-executable instructions for performing a method comprising:

producing at least one first classification result using one or more first classifiers and one or more first data sources;

producing a first decision using the at least one first classification result and a first set of one or more respective weights respectively associated with the one or more first classifiers;

selecting at least one second classifier if the first decision has a first prescribed value;

producing at least one second classification result using the one or more first classifiers and the at least one second classifier and the one or more first data sources and at least one second data source; and

producing a second decision using the at least one second classification result and a second set of one or more respective weights respectively associated with the one or more first classifiers and with the at least one second classifier.

19. The computer readable medium of claim 18, wherein the method further comprises:

selecting at least one third classifier if the first decision has a second prescribed value;

producing at least one third classification result using the one or more first classifiers and the at least one third classifier and the one or more first data sources and at least one third data source; and

producing a third decision using the at least one second classification result and a third set of one or more respective weights respectively associated with the one or more first classifiers and with the at least one third classifier.

20. The computer readable medium of claim 19, wherein the method further comprises:

selecting at least one fourth classifier if the third decision has a first prescribed value;

producing at least one fourth classification result using the one or more first classifiers, the at least one third and the at least one fourth classifiers and the one or more first and third data sources and at least one fourth data source;

producing a fourth decision using the at least one fourth classification result and a fourth set of one or more respective weights respectively associated with the one or more first classifiers and with the at least one third and the at least one fourth classifiers.