Learning Device, Learning Method, and Program

• US 20100010948A1
• Filed: 07/07/2009
• Published: 01/14/2010
• Est. Priority Date: 07/09/2008
• Status: Active Grant

• Current Assignee: Sony Corporation (Sony Group Corp.)
• Original Assignee: Sony Corporation (Sony Group Corp.)
• Inventors: ITO, Masato, AOYAMA, Kazumi, NODA, Kuniaki
• Granted Patent: US 8,306,930 B2
• Time in Patent Office: Days
• US Class Current: 706/20
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/08 Learning methods

Abstract

A learning device includes: a plurality of learning modules, each of which performs update learning to update a plurality of model parameters of a pattern learning model that learns a pattern using input data; model parameter sharing means for causing two or more learning modules from among the plurality of learning modules to share the model parameters; and sharing strength updating means for updating sharing strengths between the learning modules so as to minimize learning errors when the plurality of model parameters are updated by the update learning.

Claims

1. A learning device comprising:

a plurality of learning modules, each of which performs update learning to update a plurality of model parameters of a pattern learning model that learns a pattern using input data;

model parameter sharing means for causing two or more learning modules from among the plurality of learning modules to share the model parameters; and

sharing strength updating means for updating sharing strengths between the learning modules so as to minimize learning errors when the plurality of model parameters are updated by the update learning.

2. The learning device according to claim 1, wherein, when the sharing strength updating means updates the sharing strength that indicates a strength of influence from a first learning module to a second learning module from among the plurality of learning modules, the sharing strength updating means updates the sharing strength using the inner product of a parameter updating vector when the second learning module learns and a model parameter differential vector obtained by subtracting the model parameters of the second learning module from the model parameters of the first learning module.

3. The learning device according to claim 1, wherein the pattern learning model is a model that learns a time-series pattern or dynamics.

4. The learning device according to claim 1, wherein the pattern learning model is an HMM, an RNN, an FNN, an SVR or an RNNPB.

5. The learning device according to claim 1, wherein the model parameter sharing means causes all or a portion of the plurality of learning modules to share the model parameters.

6. The learning device according to claim 1, wherein the model parameter sharing means causes two or more learning modules from among the plurality of learning modules to share all or a portion of the plurality of model parameters.

7. The learning device according to claim 1, wherein the model parameter sharing means corrects the model parameters updated by each of the two or more learning modules using a weight average value of the model parameters updated respectively by the two or more learning modules to thereby cause the two or more learning modules to share the model parameters updated respectively by the two or more learning modules.

8. A learning method comprising the steps of:

performing update learning to update a plurality of model parameters of a pattern learning model that learns a pattern using input data in each of a plurality of learning modules;

causing two or more learning modules from among the plurality of learning modules to share the model parameters; and

updating sharing strengths between the learning modules so as to minimize learning errors when the plurality of model parameters are updated by the update learning.

9. A program for causing a computer to function as:

a plurality of learning modules, each of which performs update learning to update a plurality of model parameters of a pattern learning model that learns a pattern using input data;

model parameter sharing means for causing two or more learning modules from among the plurality of learning modules to share the model parameters; and

sharing strength updating means for updating sharing strengths between the learning modules so as to minimize learning errors when the plurality of model parameters are updated by the update learning.

10. A learning device comprising:

a plurality of learning modules, each of which performs update learning to update a plurality of model parameters of a pattern learning model that learns a pattern using input data;

a model parameter sharing unit that causes two or more learning modules from among the plurality of learning modules to share the model parameters; and

a sharing strength updating unit that updates sharing strengths between the learning modules so as to minimize learning errors when the plurality of model parameters are updated by the update learning.