Intelligence information processing system

• US 5,257,343 A
• Filed: 08/14/1991
• Issued: 10/26/1993
• Est. Priority Date: 09/26/1990
• Status: Expired due to Term

• Current Assignee: Mitsubishi Denki Kabushiki Kaisha (Mitsubishi Electric Corporation)
• Original Assignee: Mitsubishi Denki Kabushiki Kaisha (Mitsubishi Electric Corporation)
• Inventors: Yamaguchi, Masahiro, Ohta, Jun, Oita, Masaya, Ohyama, Nagaaki, Kyuma, Kazuo, Tai, Shuichi
• Primary Examiner(s): Fleming, Michael R.
• Assistant Examiner(s): Downs, Robert W.
• Application Number:
  US07/744,991
• Time in Patent Office: 804 Days
• Field of Search: 395/23, 395/22
• US Class Current: 706/18
• CPC Class Codes: G06N 3/02 Neural networks
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks

Abstract

An intelligence information processing system is composed of an associative memory and a serial processing-type computer. Input pattern information is associated with the associative memory, and pattern recognition based on the computer evaluates an associative output. In accordance with this evaluation, an associative and restrictive condition is repeatedly added to the energy function of a neural network constituting the associative memory, thereby converging the associative output on a stable state of the energy. The converged associative output is verified with intelligence information stored in a computer memory. The associative and restrictive condition is again repeatedly added to the energy function in accordance with the verification so as to produce an output from the system.

Claims

1. An intelligence information processing system using a serial processing-type computer, comprising:

associative memory means comprising a neutral network responsive to input pattern information for providing an associated pattern output;

pattern recognition means disposed in the serial computer and responsive to the associative memory means for evaluating whether the associated pattern output corresponds to correct pattern information;

means, responsive to an evaluation that the associated output pattern does not correspond to correct pattern information, for adding an associative and restrictive condition to an energy function of said neural network and converging said associative pattern output on said correct pattern information;

memory means for storing intelligence information;

logical processing means responsive to the converged associative pattern output for verifying the converged associated pattern output with said intelligence information; and

means, responsive to the logical processing means when the converged associative pattern output differs from the intelligence information, for repeatedly adding an associative and restrictive condition to said energy function of said neural network so as to match said converged associative pattern output with said intelligence information.

2. A system according to claim 1, wherein said pattern recognition means is a character pattern recognition means.

3. A system according to claim 1, wherein said pattern recognition means is a speech pattern recognition means.

4. A system according to claim 1, wherein said pattern recognition means is an image pattern recognition means.

5. The system of claim 1 wherein the neural network is an optical neural network.

6. The system of claim 1 wherein the pattern recognition means provides a "process-impossible" output in response to a predetermined number of evaluations that the associated pattern output does not correspond to correct pattern information.

7. The system of claim 1 wherein the logical processing means provides a "process-impossible" output when the converged associative pattern output differs from the intelligence information a predetermined number of times.

8. A method for processing an input sequence of elements for comparison with a predetermined set of sequences of a predetermined set of elements, comprising the steps of:

providing the input sequence of elements to an input of an associative memory comprising a neural network;

obtaining an associated output sequence of elements from the associative memory;

evaluating each element of the associated output sequence of elements to determine whether each element is a member of the predetermined set of elements;

improving the energy function of the neural network in response to a determined that an element is not in the predetermined set of elements;

repeating the steps of providing and evaluating in response to a determination that an element is not in the predetermined set, providing the elements as the input to the associative memory;

evaluating the associated output sequence of elements to determine whether the associated output sequence is a member of the predetermined set of sequences;

improving the energy function of the neural network in response to a determination that the associated sequence is not in the predetermined set of sequences;

repeating the steps of providing and evaluating in response to a determination that the associated sequence is not in the predetermined set of sequences, providing the input sequence of elements to the associated memory; and

wherein the step of providing is more particularly the step of providing the input sequence of elements to an optical neural network.

9. A method for processing an input sequence of elements for comparison with a predetermined set of sequences of a predetermined set of elements, comprising the steps of:

providing the input sequence of elements to an input of an associative memory comprising a neural network;

obtaining an associated output sequence of elements from the associative memory;

evaluating each element of the associated output sequence of elements to determine whether each element is a member of the predetermined set of elements;

improving the memory function of the neural network in response to a determination that an element is not in the predetermined set of elements;

repeating the steps of providing and evaluating in response to a determination that an element is not in the predetermined set, providing the element as the input to the associative memory;

evaluating the associated output sequence of elements to determine whether the associated output sequence is a member of the predetermined set of sequences;

improving the energy function of the neural network in response to a determination that the associated sequence is not in the predetermined set of sequences;

repeating the steps of providing and evaluating in response to a determination that the associated sequence is not in the predetermined set of sequences, providing the input sequence of elements to the associated memory; and

providing the predetermined set of sequences and the predetermined set of elements in a memory in a serial-type processing computer.

10. A system for processing information comprising an input sequence of elements, for comparing the input sequence of elements to a predetermined set of sequences of a predetermined set of elements, the system comprising:

an associative memory which stores the predetermined set of elements in a neural network having an input for receiving the input sequence of elements and output which provides an associated output in response to an input;

a serial processing-type computer connected to the output of the associative memory and including;

a pattern recognition unit, which stores the predetermined set of elements and receives the output of the associative memory, for determining whether each element of the associated output is a member of the predetermined set of elements;

a memory having stored therein the predetermined set of sequences;

a logical processing unit which receives the output of the associative memory, for determining whether the associated output is an element of the predetermined set of sequences by comparing the associated output to the sequences stored in the memory;

the pattern recognition unit providing a feedback signal to the associated memory, the feedback signal comprising information for correcting an energy function of the neural network and a reassociation command for instructing the associative memory to provide another associated output for the input sequence of elements;

the logical processing unit providing a feedback signal to the associative memory, the feedback signal comprising information for correcting an energy function of the neural network and a reassociation command for instructing the associative memory to provide another associated output for the input sequence of elements.

11. The system of claim 10 wherein the pattern recognition unit, after processing the same sequence of elements a predetermined number of times, provides a "process-impossible" output.

12. The system of claim 10 wherein the logical processing unit, after processing the same sequence of elements a predetermined number of times, provides a "process-impossible" output.