Training a self-learning network using interpolated input sets based on a target output

• US 9,342,793 B2
• Filed: 08/31/2010
• Issued: 05/17/2016
• Est. Priority Date: 08/31/2010
• Status: Active Grant

• Current Assignee: Red Hat, Inc. (International Business Machines Corporation)
• Original Assignee: Red Hat, Inc. (International Business Machines Corporation)
• Inventors: Williamson, Eric
• Primary Examiner(s): Hill, Stanley K
• Assistant Examiner(s): Bharadwaj, Kalpana
• Application Number:
  US12/872,935
• Publication Number:
  US 20120054131A1
• Time in Patent Office: 2,086 Days
• Field of Search: 706/12
• US Class Current: 1/1
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/08 Learning methods

Abstract

Embodiments relate to systems and methods for training a self-learning network using interpolated input sets based on a target output. A database management system can store sets of operational data, such as financial, medical, climate or other information. A user can input or access a set of target data, representing an output which a user wishes to be generated from an interpolated set of input data. The interpolation engine can generate a conformal interpolation function and input sets that map to the set of target output data. After interpolation, the interpolation engine can transmit the interpolated inputs, along with the set of target output data and other information, to a self-learning network such as a neural or fuzzy logic network. The self-learning network can be trained to converge to the target output based on the interpolated input results as generated by the interpolation engine, thus reproducing the desired interpolation function.

Claims

1. A method comprising:

accessing a set of target output data and a set of combined input data, the set of combined input data comprising a subset of predetermined input data, and a subset of interpolated input data;

receiving the set of combined input data in a self-learning network hosted on a self-learning host machine;

applying a set of weights to the set of combined input data;

revising the set of weights in view of a difference between an output generated by the self-learning network and the set of target output data;

determining that the self-learning network generates successive outputs that converge to the set of target output data in response to an application of the set of weights, as revised; and

training, by a processor and in view of the determining, the self-learning network to generate the set of target output data in conjunction with the set of weights, as revised.

2. The method of claim 1, wherein the set of combined input data comprises at least one of a set of financial data, a set of medical data, a set of demographic data, a set of engineering data, a set of network operations data, or a set of geographic data.

3. The method of claim 1, wherein the subset of interpolated input data comprises a series of subsets of interpolated input data.

4. The method of claim 3, wherein the series of subsets of interpolated input data comprises subsets of interpolated input data generated in view of user input.

5. The method of claim 3, wherein the training the self-learning network comprises training the self-learning network in view of each of the series of subsets of interpolated input data in relation to the set of target output data.

6. The method of claim 1, wherein the self-learning host machine is remote from a client machine hosting an interpolation engine to generate the set of combined input data.

7. The method of claim 1, further comprising re-training the self-learning network in view of an additional set of target output data.

8. The method of claim 1, further comprising uploading the trained self-learning network to a remote interpolation engine.

9. The method of claim 1, wherein at least one of the set of target output data or the subset of predetermined input data are accessed from a database management system remote to a self-learning host machine.

10. The method of claim 1, wherein the self-learning network comprises at least one of a neural network or a fuzzy logic network.

11. The method of claim 10, wherein training the self-learning network comprises generating the set of weights to apply to the set of combined input data to generate the set of target output data.

12. A system comprising:

an interface to a database to store a set of target output data and a set of combined input data, the set of combined input data comprising a subset of predetermined input data, and a subset of interpolated input data; and

a processor to communicate with the database via the interface and to;

receive the set of combined input data in a self-learning network hosted on a self-learning host machine,

apply a set of weights to the set of combined input data,

revise the set of weights in view of a difference between an output generated by the self-learning network and the set of target output data,

compute a determination that the self-learning network generates successive outputs that converge to the set of target output data in response to an application of the set of weights, as revised, and

train the self-learning network to generate the set of target output data in conjunction with the set of weights, as revised and in view of the determination.

13. The system of claim 12, wherein the set of combined input data comprises at least one of a set of financial data, a set of medical data, a set of demographic data, a set of engineering data, a set of network operations data, or a set of geographic data.

14. The system of claim 12, wherein the subset of interpolated input data comprises a series of subsets of interpolated input data.

15. The system of claim 14, wherein the series of subsets of interpolated input data comprises subsets of interpolated input data generated in view of user input.

16. The system of claim 14, wherein to train the self-learning network is to train the self-learning network in view of each of the series of subsets of interpolated input data in relation to the set of target output data.

17. The system of claim 12, wherein the processor is further to re-train the self-learning network in view of an additional set of target output data.

18. The system of claim 12, wherein the processor is further to upload the trained self-learning network to a remote interpolation engine.

19. The system of claim 12, wherein the self-learning network comprises at least one of a neural network or a fuzzy logic network.

20. The system of claim 19, wherein to train the self-learning network is to generate the set of weights to apply to the set of combined input data to generate the set of target output data.