AUTOMATIC CORRECTION OF INDIRECT BIAS IN MACHINE LEARNING MODELS

• US 20200134493A1
• Filed: 10/31/2018
• Published: 04/30/2020
• Est. Priority Date: 10/31/2018
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Bhide, Manish, Puri, Ruchir, Chamarthy, Ravi Chandra
• Granted Patent: US 11,068,797 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 9/547 Remote procedure calls [RPC]; Web services
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/048 Fuzzy inferencing

Abstract

Systems and methods for detecting indirect bias in machine learning models are provided. A computer-implemented method includes: receiving, by a computer device, a user request to detect transitive bias in a machine learning model; determining, by the computer device, correlations of attributes of neighboring data not included in a dataset of the machine learning model; ranking, by the computer device, the attributes based on the determined correlations; and returning, by the computer device, a list of the ranked attributes to a user that generated the user request.

Claims

1. A computer-implemented method comprising:

receiving, by a computer device, a user request to detect transitive bias in a machine learning model;

determining, by the computer device, correlations of attributes of neighboring data not included in a dataset of the machine learning model;

ranking, by the computer device, the attributes based on the determined correlations; and

returning, by the computer device, a list of the ranked attributes to a user that generated the user request.

2. The computer-implemented method of claim 1, wherein:

the computer device is a server;

the server receives the user request from an enterprise device via a callback application program interface (API); and

the server returns the list of the ranked attributes to the enterprise device via the callback API.

3. The computer-implemented method of claim 1, wherein the dataset of the machine learning model comprises training data and run time data.

4. The computer-implemented method of claim 3, further comprising the computer device determining the neighboring data prior to the determining the correlations of the attributes.

5. The computer-implemented method of claim 1, wherein the user request comprises:

an identification of the machine learning model; and

an identification of feature vectors of the machine learning model.

6. The computer-implemented method of claim 1, wherein the user request comprises an indication to perform run time bias detection of the machine learning model.

7. The computer-implemented method of claim 1, wherein the user request comprises an indication to perform design time bias detection of the machine learning model.

8. A computer program product, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computing device to cause the computer device to:

receive a user request to detect hidden bias in a machine learning model;

classify data included in a dataset of the machine learning model based on the user request;

obtain auxiliary information based on the classifying;

determine correlations of attributes of the auxiliary information;

rank the attributes based on the determined correlations; and

return a list of the ranked attributes to a user that generated the user request.

9. The computer program product of claim 8, wherein the computer device is a server that receives the user request from an enterprise device via a callback application program interface (API).

10. The computer program product of claim 9, wherein the server returns the list of the ranked attributes to the enterprise device via the callback API.

11. The computer program product of claim 8, wherein the dataset of the machine learning model comprises training data and run time data.

12. The computer program product of claim 8, wherein the user request comprises an identification of the machine learning model.

13. The computer program product of claim 12, wherein the user request comprises an indication to perform run time bias detection of the machine learning model.

14. The computer program product of claim 12, wherein the user request comprises an indication to perform design time bias detection of the machine learning model.

15. A system comprising:

a processor, a computer readable memory, and a computer readable storage medium;

program instructions to receive a user request to detect hidden bias in a machine learning model;

program instructions to classify data included in a dataset of the machine learning model;

program instructions to obtain auxiliary information based on the classifying;

program instructions to determine correlations of attributes of the auxiliary information;

program instructions to rank the attributes based on the determined correlations; and

program instructions to return a list of the ranked attributes to a user that generated the user request,

wherein the program instructions are stored on the computer readable storage medium for execution by the processor via the computer readable memory.

16. The system of claim 15, wherein:

the processor is included in a server;

the server receives the user request from an enterprise device via a callback application program interface (API); and

the server returns the list of the ranked attributes to the enterprise device via the callback API.

17. The system of claim 16, wherein the dataset of the machine learning model comprises training data and run time data.

18. The system of claim 17, wherein the user request comprises an identification of the machine learning model.

19. The system of claim 18, wherein the user request comprises an indication to perform run time bias detection of the machine learning model.

20. The system of claim 18, wherein the user request comprises an indication to perform design time bias detection of the machine learning model.