Method and apparatus for input classification using a neural network

• US 5,974,404 A
• Filed: 06/24/1997
• Issued: 10/26/1999
• Est. Priority Date: 06/19/1992
• Status: Expired due to Term

• Current Assignee: United Parcel Service Of America Incorporated (United Parcel Service Incorporated)
• Original Assignee: United Parcel Service Of America Incorporated (United Parcel Service Incorporated)
• Inventors: Moed, Michael C., Lee, Chih-Ping
• Primary Examiner(s): Downs, Robert W.
• Application Number:
  US08/881,133
• Time in Patent Office: 854 Days
• Field of Search: 706/25, 706/20, 706/6, 382/155, 382/156, 382/157, 382/158, 382/159, 382/160, 382/161
• US Class Current: 706/25
• CPC Class Codes: G06F 18/23213 with fixed number of clusters, e.g. K-means clustering
• CPC Class Codes: G06F 18/24137 Distances to cluster centroïds
• CPC Class Codes: G06F 18/2414 Smoothing the distance, e.g. radial basis function networks [RBFN]
• CPC Class Codes: G06F 18/24143 Distances to neighbourhood prototypes, e.g. restricted Coulomb energy networks [RCEN]
• CPC Class Codes: G06V 30/19107 Clustering techniques
• CPC Class Codes: G06V 30/19173 Classification techniques

Abstract

The present invention is a classification method and apparatus for classifying an input into one of a plurality of possible outputs. The invention is also a method and apparatus for adjusting a neuron encompassing a plurality of feature vectors. The invention characterizes the spatial distribution of the feature vectors. The invention then spatially adjusts the neuron in accordance with that characterization.

Claims

1. A method for refining a neuron to efficiently encompass a plurality of feature vectors, wherein feature vectors are neither added to nor removed from said neuron, including the steps of:

(a) characterizing the spatial distribution of said feature vectors; and

(b) spatially adjusting said neuron in accordance with the characterization of step (a).

2. The method of claim 1, wherein said neuron comprises a boundary defined by two or more neuron areas of different lengths.

3. The method of claim 1, wherein said neuron comprises a center point, and wherein step (b) comprises the step of translating said center point of said neuron toward the mean of said spatial distribution.

4. The method of claim 3, wherein said translated neuron comprises a boundary defined by two or more neuron axes, wherein said translated neuron does not encompass all of said feature vectors, and wherein step (b) further comprises the step of changing the length of at least one of said neuron axes to encompass all of said feature vectors.

5. An apparatus for refining a neuron to efficiently encompass a plurality of feature vectors, wherein said apparatus neither adds nor removes feature vectors to or from said neuron, comprising:

(a) characterizing means for characterizing the spatial distribution of said feature vectors; and

(b) adjusting means for spatially adjusting said neuron in accordance with the characterization of step (a).

6. The apparatus of claim 5, wherein said neuron comprises a boundary defined by two or more neuron areas of different lengths.

7. The apparatus of claim 5, wherein said neuron comprises a center point, and wherein said adjusting means translates said center point of said neuron toward the mean of said spatial distribution.

8. The apparatus of claim 7, wherein said translated neuron comprises a boundary defined by two or more neuron axes, wherein said translated neuron does not encompass all of said feature vectors, and wherein said adjusting means changes the length of at least one of said neuron axes to encompass all of said feature vectors.