Visualization and self-organization of multidimensional data through equalized orthogonal mapping

• US 6,907,412 B2
• Filed: 03/23/2001
• Issued: 06/14/2005
• Est. Priority Date: 09/29/1995
• Status: Expired due to Term

• Current Assignee: CA, Inc. (d/b/a CA Technologies) (Broadcom, Inc.)
• Original Assignee: Computer Associates Think Inc. (Broadcom, Inc.)
• Inventors: Meng, Zhuo, Pao, Yoh-Han
• Primary Examiner(s): STARKS, WILBERT L
• Application Number:
  US09/816,909
• Publication Number:
  US 20010032198A1
• Time in Patent Office: 1,544 Days
• Field of Search: 706/16, 706/2, 382/118, 700/83, 704/9, 704/254, 364/188, 395/794
• US Class Current: 706/16
• CPC Class Codes: G06F 18/213 Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
• CPC Class Codes: G06F 18/21355 nonlinear criteria, e.g. embedding a manifold in a Euclidean space
• CPC Class Codes: G06F 18/2137 based on criteria of topology preservation, e.g. multidimensional scaling or self-organising maps

Abstract

The subject system provides reduced-dimension mapping of pattern data. Mapping is applied through conventional single-hidden-layer feed-forward neural network with non-linear neurons. According to one aspect of the present invention, the system functions to equalize and orthogonalize lower dimensional output signals by reducing the covariance matrix of the output signals to the form of a diagonal matrix or constant times the identity matrix. The present invention allows for visualization of large bodies of complex multidimensional data in a relatively “topologically correct” low-dimension approximation, to reduce randomness associated with other methods of similar purposes, and to keep the mapping computationally efficient at the same time.

Claims

1. A system for visualizing multi-dimensional pattern data reduced to a lower dimension representation, comprising:

a neural network having an input layer and an other layer, wherein a number of nodes in the other layer is less than a number of input nodes in the input layer, and the other layer supplies an output signal corresponding to multi-dimensional pattern data received by the input layer; and

a training module for the neural network, wherein the training module includes means for equalizing and orthogonalizing the output signal of the other layer.

2. The system of claim 1, wherein the training module equalizes and orthogonalizes the output signal of the other layer by constraining values of a covariance matrix of the output signal.

3. The system of claim 1, wherein the training module equalizes and orthogonalizes the output signal of the other layer by reducing a covariance matrix of the output signal to a form of a diagonal matrix.

4. The system of claim 1, wherein output data is collected from the neural network, and a two-dimensional map of the output data is displayed.

5. The system of claim 1, wherein output data is collected from the neural network, and a plurality of two-dimensional maps of the output data are displayed.

6. The system of claim 1, wherein the lower-dimension representation is a three dimensional display.

7. The system of claim 1, wherein the training module performs self-supervised training.

8. The system of claim 1, wherein the neural network is self organizing.

9. The system of claim 1, wherein nodes in the other layer are non-linear.

10. The system of claim 1, wherein the other layer comprises an output layer.

11. A method for visualizing multi-dimensional pattern data reduced to a lower dimension representation, comprising:

providing a neural network having an input layer and an other layer, wherein a number of nodes in the other layer is less than a number of input nodes in the input layer, and the other layer supplies an output signal corresponding to multi-dimensional pattern data received by the input layer; and

training the neural network to equalize and orthogonalize the output signal of the other layer.

12. The method of claim 11, wherein the output signal of the other layer is equalized and orthogonalized by constraining values of a covariance matrix of the output signal.

13. The method of claim 11, wherein the output signal of the other layer is equalized and orthogonalized by reducing a covariance matrix of the output signal to a form of a diagonal matrix.

14. The method of claim 11, further comprising collecting output data from the neural network, and displaying a two-dimensional map of the output data.

15. The method of claim 11, further comprising collecting output data from the neural network, and displaying a plurality of two-dimensional maps of the output data.

16. The method of claim 11, wherein the lower-dimension representation is a three dimensional display.

17. The method of claim 11, wherein the training is self supervised training.

18. A computer system, comprising:

a processor; and

a program storage device readable by the computer system, tangibly embodying a program of instructions executable by the processor to perform the method claimed in claim 11.

19. A program storage device readable by a machine, tangibly embodying a program of instructions executable by the machine to perform the method claimed in claim 11.

20. A computer data signal transmitted in one or more segments in a transmission medium which embodies instructions executable by a computer to perform the method claimed in claim 11.