Stochastic encoder/decoder/predictor

• US 6,424,956 B1
• Filed: 03/18/1999
• Issued: 07/23/2002
• Est. Priority Date: 07/13/1993
• Status: Expired due to Fees

• Current Assignee: IPU Power Management, LLC
• Original Assignee: Paul J. Werbos
• Inventors: Werbos, Paul J.
• Primary Examiner(s): Black, Thomas
• Assistant Examiner(s): STARKS, WILBERT L
• Application Number:
  US09/271,439
• Time in Patent Office: 1,223 Days
• Field of Search: 706/21, 706/2, 706/37, 706/41
• US Class Current: 706/2
• CPC Class Codes: G06N 7/04 Physical realisation

Abstract

An artificial intelligence system is provided which makes use of a dual subroutine to adapt weights. Elastic Fuzzy Logic (“ELF”) System is provided in which classical neural network learning techniques are combined with fuzzy logic techniques in order to accomplish artificial intelligence tasks such as pattern recognition, expert cloning and trajectory control. The system may be implemented in a computer provided with multiplier means and storage means for storing a vector of weights to be used as multiplier factors in an apparatus for fuzzy control.

Claims

1. An apparatus for neural network artificial intelligence in a system having target outputs and target deviators and a series of weights, comprising:

a central processing unit;

an input device for entering data to said central processing unit; and

an output device for producing an output from said central processing unit, said input device providing a series of inputs, desired outputs, and desired output deviators to said CPU, and said processing unit computing an error function using a dual subroutine and adapting the weights in response to the error function wherein the apparatus is used as a controller for a physical machine.

2. The apparatus as claimed in claim 1, wherein the apparatus is used as a controller for an airplane.

3. The apparatus as claimed in claim 1, wherein the dual subroutines are configured to propagates derivatives through a differentiable network.

4. The apparatus as claimed in claim 3, wherein the derivatives are used as inputs to an adaptable network.

5. The apparatus as claimed in claim 3, wherein the derivatives are used as inputs to an adaptable network using steepest descent learning.

6. The apparatus as claimed in claim 3, wherein the derivatives are used as inputs to an adaptable network using conjugate gradient learning.

7. The apparatus as claimed in claim 1, wherein the dual subroutine comprises software.

8. The apparatus as claimed in claim 1, wherein the error function is calculated using a chain rule for ordered derivatives.

9. The apparatus as claimed in claim 1, wherein the dual subroutine comprises hardware.

10. The apparatus as claimed in claim 1, further comprising a critic network configured to be adapted.

11. An apparatus for artificial intelligence, comprising:

a central processing unit;

an input device for entering data to said central processing unit; and

an output device for producing an output, TARGET, from said central processing unit, said input device providing values of derivatives F0_zi, said central processing unit performing a calculation of the form $F_{\mathrm{zi}}={\mathrm{F0\_z}}_i+\underset{j=i+1}{\overset{N-1}{\sum <br><br>}}<br><br>F_{\mathrm{zj}}*\left(\partial <br><br>z_j/\partial <br><br>z_i\right),$

wherein F0_z_iΔ

(∂

TARGET/∂

z_i) where m+1≦

i≦

N−

1 for a module that calculates

12. The apparatus as claimed in claim 11, wherein the apparatus is used as a controller for an airplane.