System and method facilitating pattern recognition

• US 20030174881A1
• Filed: 03/15/2002
• Published: 09/18/2003
• Est. Priority Date: 03/15/2002
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Platt, John C., Simard, Patrice Y., Steinkraus, David Willard
• Granted Patent: US 7,016,529 B2
• Time in Patent Office: Days
• US Class Current: 382/159
• CPC Class Codes: G06V 10/454 Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
• CPC Class Codes: G06V 30/287 of Kanji, Hiragana or Katakana characters
• CPC Class Codes: G06V 30/36 Matching; Classification

Abstract

A system and method facilitating pattern recognition is provided. The invention includes a pattern recognition system having a convolutional neural network employing feature extraction layer(s) and classifier layer(s). The feature extraction layer(s) comprises convolutional layers and the classifier layer(s) comprises fully connected layers. The pattern recognition system can be trained utilizing a calculated cross entropy error. The calculated cross entropy error is utilized to update trainable parameters of the pattern recognition system.

Claims

1. A pattern recognition system, comprising:

at least one convolutional layer that receives a non-temporal input pattern, the convolutional layer comprising a plurality of feature maps, the feature map receiving at least a portion of the non-temporal input pattern, the feature map including trainable parameters, the at least one convolutional layer providing outputs associated with features extracted from the non-temporal input pattern; and

,

at least one fully connected layer that receives outputs from the at least one convolutional layer, the at least one fully connected layer classifying the features extracted by the at least one convolutional layer, the at least one fully connected layer providing a plurality of outputs, the output comprising a probability associated with a class, the pattern recognition system trained utilizing cross entropy error minimization.

2. The pattern recognition system of claim 1, the trainable parameters for the feature map being equal.

3. The pattern recognition system of claim 1, cross entropy error minimization being based, at least in part, upon the following equation:

4. The pattern recognition system of claim 1 trained using a stochastic gradient descent algorithm.

5. The pattern recognition system of claim 1 trained to recognize one stroke Asian characters.

6. The pattern recognition system of claim 1 trained to recognize two stroke Asian characters.

7. The pattern recognition system of claim 1 trained to recognize one stroke Japanese characters.

8. The pattern recognition system of claim 1 trained to recognize two stroke Japanese characters.

9. The pattern recognition system of claim 1 trained to a plurality of ASCII characters.

10. A pattern recognition system, comprising:

a first convolutional layer that receives a bitmap input pattern, the first convolutional layer comprising a plurality of first feature maps, the first feature map receiving at least a portion of the bitmap input pattern, the first feature map including first trainable parameters, the first feature maps providing outputs associated with first features extracted from the bitmap input pattern;

a second convolutional layer that receives the outputs of the first feature maps, the second convolutional layer comprising a plurality of second feature maps, the second feature map receiving at least a portion of the outputs of the first feature maps, the second feature map including second trainable parameters, the second feature maps providing outputs associated with second features;

a first fully connected layer that classifies the outputs of the second feature maps, the first fully connected layer providing outputs;

a second fully connected layer that classifies the outputs of the first fully connected layer, the second fully connected layer providing a plurality of outputs, the output comprising a probability associated with a class, the pattern recognition system trained utilizing cross entropy error minimization.

11. The pattern recognition system of claim 10, the first convolutional layer comprising five first feature maps.

12. The pattern recognition system of claim 11, the first feature map receiving 156 inputs.

13. The pattern recognition system of claim 10, the second convolutional layer comprising fifty second feature maps.

14. The pattern recognition system of claim 13, the second feature map receiving 25 inputs.

15. The pattern recognition system of claim 10, the first fully connected layer comprising one hundred hidden units.

16. The pattern recognition system of claim 10, the first trainable parameters for a first feature map being equal.

17. The pattern recognition system of claim 10, cross entropy error minimization being based, at least in part, upon the following equation:

18. The pattern recognition system of claim 10 trained using stochastic gradient descent algorithm.

19. The pattern recognition system of claim 10 trained to recognize one stroke Asian characters.

20. The pattern recognition system of claim 10 trained to recognize two stroke Asian characters.

21. The pattern recognition system of claim 10 trained to recognize one stroke Japanese characters.

22. The pattern recognition system of claim 10 trained to recognize two stroke Japanese characters.

23. The pattern recognition system of claim 10 trained to a plurality of ASCII characters.

24. A document scanner employing the pattern recognition system of claim 10.

25. An optical character recognition system employing the pattern recognition system of claim 10.

26. A personal digital assistant employing the pattern recognition system of claim 10.

27. A tablet personal computer employing the pattern recognition system of claim 10.

28. A system for training a pattern recognition system, comprising:

a pattern recognition system comprising a convolutional neural network that receives a non-temporal input pattern and provides a plurality of class probability outputs;

a cross entropy error calculator that calculates a cross entropy error based, at least in part, upon the plurality of class probability outputs and training class information;

a back propagation gradient descent component that utilizes a stochastic descent algorithm to update trainable parameters of the pattern recognition system based, at least in part, upon the cross entropy error; and

a trainable parameter update component that updates the trainable parameters of the pattern recognition system.

29. The system of claim 28, the back propagation gradient descent component utilizing the following equation in updating the trainable parameters:

30. A method for training a pattern recognition system, comprising:

performing pattern recognition on a training pattern utilizing a pattern recognition system based, at least in part, upon a convolutional neural network having a set of trainable parameters;

providing a plurality of class probability outputs based on the training pattern;

calculating a cross entropy error based, at least in part, upon the plurality of class probability outputs and information associated with the training pattern; and

,

updating the set of trainable parameters based, at least in part, upon a gradient descent algorithm utilizing the calculated entropy error.

31. A data packet transmitted between two or more computer components that facilitates training a pattern recognition system, the data packet comprising:

a data field comprising a set of trainable parameters for a pattern recognition system based, at least in part, upon a convolutional neural network, the set of trainable parameters updated based, at least in part, upon a gradient descent algorithm utilizing a calculated entropy error.

32. A computer readable medium storing computer executable components of a system facilitating training of a pattern recognition, comprising:

a pattern recognition component comprising a convolutional neural network that receives a non-temporal input pattern and provides a plurality of class probability outputs;

a cross entropy error calculator component that calculates a cross entropy error based, at least in part, upon the plurality of class probability outputs and training class information;

a back propagation gradient descent component that utilizes a stochastic gradient descent algorithm to update trainable parameters of the pattern recognition system based, at least in part, upon the cross entropy error; and

a trainable parameter update component that updates the trainable parameters of the pattern recognition system.

33. A training system for a pattern recognition system, comprising:

means for inputting a non-temporal pattern;

means for performing pattern recognition utilizing a convolutional neural network that receives the non-temporal input pattern and provides a plurality of class probability outputs;

mean for calculating a cross entropy error based, at least in part, upon the plurality of class probability outputs and training class information; and

,

means for updating trainable parameters of the means for performing pattern recognition, the means for updating utilizing a stochastic gradient descent algorithm to update the trainable parameters of the pattern recognition system based, at least in part, upon the cross entropy error.