Method and System for Invariant Pattern Recognition

• US 20150227849A1
• Filed: 12/07/2010
• Published: 08/13/2015
• Est. Priority Date: 12/07/2009
• Status: Active Grant

• Current Assignee: Verizon Patent and Licensing Incorporated (Verizon Communications Inc.)
• Original Assignee: Yahoo! Inc. (Apollo Global Management, Inc.)
• Inventors: Jaros, Robert G., Osindero, Simon Kayode
• Granted Patent: US 9,361,586 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/02 Neural networks
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

An adaptive pattern recognition system optimizes an invariance objective and an input fidelity objective to accurately recognize input patterns in the presence of arbitrary input transformations. A fixed state or value of a feature output can nonlinearly reconstruct or generate multiple spatially distant input patterns and respond similarly to multiple spatially distant input patterns, while preserving the ability to efficiently evaluate the input fidelity objective. Exemplary networks, including a novel factorization of a third-order Boltzmann machine, exhibit multilayered, unsupervised learning of arbitrary transformations, and learn rich, complex features even in the absence of labeled data. These features are then used to classify unknown input patterns, to perform dimensionality reduction or compression,

Claims

1. (canceled)

9. (canceled)

21. A method to facilitate pattern recognition comprising:

processing a set of vectors into feature vectors;

wherein the processing comprises employing machine learning component adjusted parameters;

wherein the machine learning component adjusted parameters are adjusted based at least in part on at least partially conflicting learning objectives including an input fidelity objective and an invariance objective, and are further adjusted at least in part by adding one or more scaled derivatives of one or more of the at least partially conflicting learning objectives to one or more parameters.

2. The method of claim 21 further comprising:

classifying test patterns based at least in part on said set of feature vectors.

3. The method of claim 21 further comprising:

performing non-linear dimensionality reduction on said set of feature vectors for visualization or analysis.

4. The method of claim 21 wherein the input fidelity objective comprises a likelihood objective.

5. The method of claim 4 wherein the invariance objective comprises invariance for transformations and/or intra class variations.

6. The method of claim 4 wherein the at least partially conflicting learning objectives comprises a temporal coherence objective.

7. The method of claim 21 wherein the input vectors comprise images.

8. The method of claim 21 wherein the input vectors comprise audio input.

22. An apparatus comprising:

a computing device;

said computing device to facilitate pattern recognition for a set of vectors;

said computing device to further process the set of vectors into feature vectors;

wherein said computing device including a machine learning component to adjust machine learning component parameters based at least in part on at least partially conflicting learning objectives including an input fidelity objective and an invariance objective, said machine learning component further to add one or more scaled derivatives of one or more of the at least partially conflicting learning objectives to one or more machine learning component parameters to adjust said machine learning component parameters.

10. The apparatus of claim 22 further comprising:

a classifier of the machine learning component to compute a prediction.

11. The apparatus of claim 22 wherein said computing device also to perform non-linear dimensionality reduction.

12. The apparatus of claim 22 wherein individual layers to comprise a Boltzmann machine.

13. The apparatus of claim 12, wherein individual Boltzmann machine layers to comprise a factored third-order Boltzmann machine.

14. The apparatus of claim 12 wherein the apparatus comprises more than one layer.

15. The apparatus of claim 14, the apparatus to process individual layers as a one layer Boltzmann machine.

16. The apparatus of claim 22 wherein the input fidelity objective comprises a likelihood objective.

17. The apparatus of claim 22 wherein the invariance objective comprises invariance for transformations and/or intra-class variations.

18. The apparatus of claim 22 wherein the at least partially conflicting learning objectives comprises a temporal coherence objective.

19. The apparatus of claim 22 wherein the input vectors include images.

20. The apparatus of claim 22 wherein the input vectors include audio input.