System and method for composite distance metric leveraging multiple expert judgments

• US 8,996,443 B2
• Filed: 09/23/2013
• Issued: 03/31/2015
• Est. Priority Date: 10/08/2010
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Ebadollahi, Shahram, Sun, Jimeng, Wang, Fei
• Primary Examiner(s): CHEN, ALAN S
• Application Number:
  US14/033,801
• Publication Number:
  US 20140025618A1
• Time in Patent Office: 554 Days
• Field of Search: None
• US Class Current: 706/48
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: G06N 5/047 Pattern matching networks; Rete networks
• CPC Class Codes: G06N 7/00 Computing arrangements based on specific mathematical models
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients
• CPC Class Codes: G16Z 99/00 Subject matter not provided for in other main groups of this subclass

Abstract

A system and method for a composite distance metric leveraging multiple expert judgments includes inputting a data distribution of multiple expert judgments stored on a computer readable storage medium. Base distance metrics are converted into neighborhoods for comparison, wherein each base distance metric represents an expert. The neighborhoods are combined to leverage the local discriminalities of all base distance metrics by applying at least one iterative process to output a composite distance metric.

Claims

1. A method for processing data, comprising:

converting base distance metrics from a plurality of sources into neighborhoods for comparison, wherein each base distance metric represents an expert; and

combining the neighborhoods to leverage the local discriminalities of all base distance metrics using a processor by applying at least one iterative process to output a composite distance metric.

2. The method as recited in claim 1, further comprising:

adjusting the composite distance metric to incorporate a new feature vector or a new base distance metric.

3. The method as recited in claim 1, wherein the neighborhoods are identified as a homogeneous neighborhood or a heterogeneous neighborhood.

4. The method as recited in claim 3, wherein:

the homogeneous neighborhood includes the nearest neighbors, based on input neighbor size, with a same input label; and

the heterogeneous neighborhood includes the nearest neighbors, based on input neighbor size, with a different input label.

5. The method as recited in claim 1, wherein the iterative process includes alternating optimization.

6. The method as recited in claim 5, wherein alternating optimization includes eigenvalue decomposition and Euclidean projection.

7. The method as recited in claim 1, wherein:

an individual similarity measure is determined for each expert; and

the composite distance metric is provided back to each expert to compare to the individual similarity measure.

8. The method as recited in claim 1, wherein a similarity measure between a first patient with a first vector and a second patient with a second vector is determined by computing a Mahalanobis distance between the first vector and the second vector.

9. The method as recited in claim 1, wherein the data distribution is a multilinear data distribution.

10. A system for processing data, comprising:

a neighborhood formulation module configured to convert base distance metrics from a plurality of sources into neighborhoods for comparison, wherein each base distance metric represents an expert; and

a neighborhood combination module, having a processor, configured to combine the neighborhoods to leverage the local discriminalities of all base distance metrics by applying at least one iterative process to output a composite distance metric.

11. The system as recited in claim 10, further comprising:

a distance update module configured to adjust the composite distance metric to incorporate a new feature vector or a new base distance metric.

12. The system as recited in claim 10, wherein the neighborhoods are identified as a homogeneous neighborhood or heterogeneous neighborhood.

13. The system as recited in claim 12, wherein:

the homogeneous neighborhood includes the nearest neighbors, based on input neighbor size, with a same input label; and

the heterogeneous neighborhood includes the nearest neighbors, based on input neighbor size, with a different input label.

14. The system as recited in claim 10, wherein the iterative process includes alternating optimization.

15. The system as recited in claim 14, wherein alternating optimization includes eigenvalue decomposition and Euclidean projection.

16. The system as recited in claim 10, wherein:

an individual similarity measure is determined for each expert; and

the composite distance metric is provided back to each expert to compare to the individual similarity measure.

17. The system as recited in claim 10, wherein a similarity measure between a first patient with a first vector and a second patient with a second vector is determined by computing a Mahalanobis distance between the first vector and the second vector.

18. The system as recited in claim 10, wherein the data distribution includes at least one of a multilinear data distribution and a nonlinear data distribution.

19. The system as recited in claim 10, wherein the data distribution is a nonlinear data distribution.

20. A computer readable storage medium comprising a computer readable program, wherein the computer readable program when executed on a computer causes the computer to perform the steps of:

converting base distance metrics from a plurality of sources into neighborhoods for comparison, wherein each base distance metric represents an expert; and

combining the neighborhoods to leverage the local discriminalities of all base distance metrics by applying at least one iterative process to output a composite distance metric.