Adaptive learning system and method

• US 7,089,217 B2
• Filed: 04/10/2001
• Issued: 08/08/2006
• Est. Priority Date: 04/10/2000
• Status: Expired due to Term

• Current Assignee: Pacific Edge Biotechnology Limited
• Original Assignee: Pacific Edge Biotechnology Limited
• Inventors: Kasabov, Nikola Kirilov
• Primary Examiner(s): Davis, George
• Application Number:
  US10/257,214
• Publication Number:
  US 20030149676A1
• Time in Patent Office: 1,946 Days
• Field of Search: 706/2, 706/14
• US Class Current: 706/12
• CPC Class Codes: G06N 3/043 based on fuzzy logic, fuzzy membership or fuzzy inference, e.g. adaptive neuro-fuzzy inference systems [ANFIS]

Abstract

A neural network module including an input layer having one or more input nodes arranged to receive input data, a rule base layer having one or more rule nodes, an output layer having one or more output nodes, and an adaptive component arranged to aggregate selected two or more rule nodes in the rule base layer based on the input data, an adaptive learning system having one or more of the neural network modules, related methods of implementing the neural network module and an adaptive learning system, and a neural network program.

Claims

1. A neural network program embodied on a computer-readable medium comprising:

an input layer comprising one or more input nodes arranged to receive input data;

a rule base layer comprising one or more rule nodes;

an output layer comprising one or more output nodes, each rule node in the rule base layer having a minimum activation threshold and each rule node arranged to be activated when input data satisfies the minimum activation threshold of the rule node; and

an adaptive component arranged to aggregate selected two or more rule nodes in the rule base layer based on the input data.

2. The neural network program as claimed in claim 1, wherein the parameters of the activation threshold of each rule node activated by input data are adjusted based on the input data.

3. The neural network program as claimed in claim 2, wherein the adaptive component is arranged to aggregate two or more rule nodes based on the magnitude of activation when activated by input data.

4. The neural network program as claimed in claim 2, wherein the parameters of activation threshold of each rule node are adjusted based at least partially on new input data.

5. The neural network program as claimed in claim 2, further comprising input data stored in a memory, the parameters of the activation threshold of each rule node arranged to be adjusted based at least partially on the stored input data.

6. The neural network program as claimed in claim 1, wherein each rule node is assigned a magnitude of activation when activated by input data.

7. The neural network program as claimed in claim 1, wherein the adaptive component is arranged to increase the minimum activation threshold of one or more rule nodes not selected for aggregation.

8. The neural network program as claimed in claim 1, wherein the parameters of the activation threshold of each rule node activated by input data are adjusted based on both the input data and desired output data.

9. The neural network program as claimed in claim 1, wherein the adaptive component is arranged to insert new rule nodes into the rule base layer.

10. The neural network program as claimed in claim 1 wherein the adaptive component is arranged to extract rules from the rule base layer.

11. The neural network program as claimed in claim 1, further comprising a fuzzy input layer comprising one or more fuzzy input nodes arranged to transform input node values for use by the rule base layer.

12. The neural network program as claimed in claim 1, further comprising a fuzzy output layer comprising one or more fuzzy output nodes arranged to transform data output from the rule base layer.

13. An adaptive learning program embodied on a computer readable medium comprising more than one neural network program each neural network program comprising:

an input layer comprising one or more input nodes arranged to receive input data;

a rule base layer comprising one or more rule nodes, each rule node in the rule base layer having a minimum activation threshold and each rule node arranged to be activated when input data satisfies the minimum activation threshold of the rule node;

an output layer comprising one or more output nodes; and

an adaptive component arranged to aggregate selected two or more rule nodes in the rule base layer based on the input data.

14. A neural network module comprising:

an input layer comprising one or more input nodes arranged to receive input data;

a rule base layer comprising one or more rule nodes, each rule node having a minimum activation threshold, each rule node configured to be activated where when input data satisfies the minimum activation threshold of the rule node;

an output layer comprising one or more output nodes; and

an adaptation component arranged to aggregate selected two or more rule nodes in the rule base layer based on the input data, and to increase the minimum activation threshold of one or more rule nodes not selected for aggregation.

15. The neural network module as claimed in claim 14, wherein the parameters of the activation threshold of each rule node activated by input data are adjusted based on the input data.

16. The neural network module as claimed in claim 14, wherein each rule node is assigned a magnitude of activation when activated by input data.

17. The neural network module as claimed in claim 16, wherein the adaptation component is configured to aggregate two or more rule nodes based on the magnitude of activation when activated by input data.

18. The neural network module as claimed in claim 14, wherein the parameters of the activation threshold of each rule node activated by input data are adjusted based on both the input data and desired output data.

19. The neural network module as claimed in claim 14, wherein the adaptation component is configured to insert new rule nodes into the rule base layer.

20. The neural network module as claimed in claim 14, wherein the adaptation component is configured to extract rules from the rule base layer.

21. The neural network module as claimed in claim 14, wherein the parameters of the activation threshold of each rule node are adjusted based at least partially on new input data.

22. The neural network module as claimed in claim 14, further comprising a memory in which is stored input data, wherein the parameters of the activation threshold of each rule node are adjusted based at least partially on the stored input data.

23. The neural network module as claimed in claim 14, further comprising a fuzzy input layer comprising one or more fuzzy input nodes arranged to transform input node values for use by the rule base layer.

24. The neural network module as claimed in claim 14, further comprising a fuzzy output layer comprising one or more fuzzy output nodes arranged to transform data output from the rule base layer.

25. An adaptive learning system comprising more than one neural network module, each neural network module comprising:

an input layer comprising one ore more input nodes arranged to receive input data;

a rule base layer comprising one or more rule nodes, each rule node having a minimum activation threshold and each rule node configured to be activated when input data satisfies the minimum activation threshold of the rule node;

an output layer comprising one or more output nodes; and

an adaptation component arranged to aggregate selected two or more rule nodes in the rule base layer based on the input data, and to increase the minimum activation threshold of one or more rule nodes not selected for aggregation.

26. A method of implementing a neural network module comprising the steps of:

maintaining in computer memory an input layer comprising one or more input nodes to receive input data;

maintaining in computer memory a rule base layer comprising one or more rule nodes;

assigning a minimum activation threshold to each rule node in the rule base layer, each rule node being activated when input data satisfies the minimum activation threshold of the rule node;

maintaining in computer memory an output layer comprising one or more output nodes;

aggregating selected two or more rule nodes in the rule base layer based on the input data; and

increasing the minimum activation threshold of one or more rule nodes not selected for aggregation.

27. The method of implementing a neural network module as claimed in claim 26, further comprising the step of adjusting the parameters of the activation threshold of each rule activated by input based on the input data.

28. The method of implementing a neural network module as claimed in claim 26, further comprising the step of assigning to each rule node a magnitude of activation when activated by input data.

29. The method of implementing a neural network module as claimed in claim 26, further comprising the step of aggregating two or more rule nodes based on the magnitude of activation when activated by input data.

30. The method of implementing a neural network module as claimed in claim 26, further comprising the step of adjusting the parameters of the activation threshold of each rule node activated by input data based on both the input data and desired output data.

31. The method of implementing a neural network module as claimed in claim 26, further comprising the step of inserting new rule nodes into the rule base layer.

32. The method of implementing a neural network module as claimed in claim 26, further comprising the step of extracting rules from the rule base layer.

33. The method of implementing a neural network module as claimed in claim 26, further comprising the step of adjusting the parameters of the activation threshold of each rule node based at least partially on new input data.

34. The method of implementing a neural network module as claimed in claim 26, further comprising the steps of maintaining in computer memory input data;

and adjusting the parameters of activation threshold of each rule node based at least partially on the stored input data.

35. The method of implementing a neural network module as claimed in claim 26, further comprising the step of maintaining in computer memory a fuzzy input layer comprising one or more fuzzy input nodes to transform input node values for use by the rule base layer.

36. The method of implementing a neural network module as claimed in claim 26, further comprising the steps of maintaining in computer memory a fuzzy output layer comprising one or more fuzzy output nodes to transform data output from the rule base layer.