Data processing system

• US 5,664,069 A
• Filed: 05/23/1995
• Issued: 09/02/1997
• Est. Priority Date: 07/10/1989
• Status: Expired due to Fees

• Current Assignee: Sharp Corporation (Hon Hai Precision Industry Co., Ltd.), YOZAN, Inc.
• Original Assignee: Sharp Corporation (Hon Hai Precision Industry Co., Ltd.), YOZAN, Inc.
• Inventors: Kumagai, Ryohei, Matsumoto, Koji, Yamamoto, Makoto, Takatori, Sunao
• Primary Examiner(s): Downs, Robert W.
• Application Number:
  US08/447,775
• Time in Patent Office: 833 Days
• Field of Search: 395/24, 395/22, 395/21, 382/158
• US Class Current: 706/31
• CPC Class Codes: G06N 3/04 Architecture, e.g. interconnection topology
• CPC Class Codes: G06N 3/063 using electronic means

Abstract

A data processing system including a plurality of neural layers characterized in that each neural layer is divided into a plurality of groups and that neurons in one of the group in one layer are connected only with neurons in corresponding group of adjacent layers, whereby independent neural cells are constructed each of which comprises corresponding groups of neurons. The minimum number of neural layers is equal to the difference between the degree of abstraction of an output data and the degree of abstraction of an input data.

Claims

1. A neural network system comprising:

an image scanning device generating lightness measurements of pixel areas of an image;

a plurality of neural cells, each cell comprising a plurality of neural layers connected in series, each layer comprising a plurality of neural processing elements, each neural processing element generating an output when a weighted sum of inputs exceeds a threshold;

wherein a first neural layer of each neural cell receives, as inputs having a given degree of abstraction, lightness measurements from a finite region within an image;

wherein processing of an input in the neural layers changes the degree of abstraction of the input; and

wherein the minimum number of neural layers is equal to the difference between the degree of abstraction of an output and the degree of abstraction of the input.

2. The system of claim 1 wherein an output of a neural processing element of a first layer of a neural cell connects to an input of each neural processing element of a second layer of the cell.

3. The system of claim 1 wherein each output of each neural processing element of a first layer of a cell connects to an input of each neural processing element of a second layer of the cell.

4. The system of claim 1 wherein neural layers of a cell are connected sequentially, and outputs of neural processing elements of each neural layer of the cell connect to inputs of neural processing elements of a subsequent neural layer of the cell.

5. The system of claim 4 wherein the inputs of neural processing elements of each neural layer have the same degree of abstraction.

6. The system of claim 1 wherein each neural processing element of a first layer of a neural cell receives, as inputs, lightness measurements from a multi-pixel region within an image.

7. The system of claim 1 wherein the neural processing elements are connected to other neural processing elements only within the same cell such that data transmission is performed only within the cell.

8. The system of claim 1 wherein processing of the inputs in each neural layer increases the degree of the abstraction of the inputs by one degree such that the number of neural layers is equal to the degree of abstraction of the output minus the degree of abstraction of the input.