METHOD AND APPARATUS FOR CONSTRUCTING, USING AND REUSING COMPONENTS AND STRUCTURES OF AN ARTIFICAL NEURAL NETWORK

• US 20150106311A1
• Filed: 10/14/2014
• Published: 04/16/2015
• Est. Priority Date: 10/16/2013
• Status: Active Grant

• Current Assignee: University of Tennessee Research Foundation (University of Tennessee)
• Original Assignee: University of Tennessee Research Foundation (University of Tennessee)
• Inventors: Birdwell, J. Douglas, Dean, Mark E., Schuman, Catherine
• Granted Patent: US 10,019,470 B2
• Time in Patent Office: Days
• US Class Current: 706/20
• CPC Class Codes: G06F 16/22 Indexing; Data structures therefor; Storage structures
• CPC Class Codes: G06N 3/02 Neural networks
• CPC Class Codes: G06N 3/049 Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
• CPC Class Codes: G06N 3/063 using electronic means
• CPC Class Codes: G06N 3/065 Analogue means
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 3/086 using evolutionary algorithms, e.g. genetic algorithms or genetic programming
• CPC Class Codes: G06N 3/10 Interfaces, programming languages or software development kits, e.g. for simulating neural networks

Abstract

A method and apparatus for constructing a neuroscience-inspired artificial neural network (NIDA) or a dynamic adaptive neural network array (DANNA) or combinations of substructures thereof comprises one of constructing a substructure of an artificial neural network for performing a subtask of the task of the artificial neural network or extracting a useful substructure based on one of activity, causality path, behavior and inputs and outputs. The method includes identifying useful substructures in artificial neural networks that may be either successful at performing a subtask or unsuccessful at performing a subtask. Successful substructures may be implanted in an artificial neural network and unsuccessful substructures may be extracted from the artificial neural network for performing the task. The method and apparatus supports constructing, using and reusing components and structures of a neuroscience-inspired artificial neural network dynamic architecture in software and a dynamic adaptive neural network array.

Claims

1. Apparatus for a neuromorphic network comprising an artificial neural network for performing a specific task, the artificial neural network comprising a plurality of substructures, each substructure for performing a sub-task of the specific task, each substructure comprising an input neuron and an output neuron, source and destination neurons connected by at least one synapse.

2. The apparatus of claim 1, each of the plurality of substructures for performing a similar subtask of the specific task.

3. The apparatus of claim 1, each of the plurality of substructures for performing a different subtask of the specific task.

4. The apparatus of claim 2 wherein each of the plurality of substructures is designed to perform the similar sub-task.

5. The apparatus of claim 1 wherein the artificial neural network comprises at least one substructure for performing a computational sub-task and at least one substructure being an affective system for performing an affective sub-task.

6. The apparatus of claim 1 further comprising a processor and an associated database of useful substructures.

7. The apparatus of claim 6 wherein said processor is adapted to extract a useful substructure for a subtask from the artificial neural network and store the useful substructure in an associated database.

8. The apparatus of claim 6 wherein said processor is adapted to implant a useful substructure for a subtask in the artificial neural network retrieved from a plurality of useful substructure stored in an associated database.

9. The apparatus of claim 6, the processor adapted to identify an unsuccessful substructure based on an activity level and extract the substructure from the artificial neural network as an unsuccessful substructure and operate the artificial neural network without the unsuccessful structure.

10. The apparatus of claim 6 wherein the processor identifies a first substructure for recognizing a first image and a second substructure for recognizing a second image different from the first image.

11. The apparatus of claim 6 wherein the processor identifies a substructure of the artificial network based on input events received by the substructure.

12. A method of constructing an artificial neural network for use in one of a control, detection and classification application comprising constructing a first substructure for performing a sub-task of a specific task and a second substructure for performing one of a similar and a different sub-task of the specific task and assembling the artificial neural network comprising the first and second substructures.

13. The method of claim 12 comprising the initial step of extracting the first substructure from a second artificial neural network as a successful substructure for performing the sub-task of the specific task.

14. The method of claim 12 wherein a plurality of circuit elements comprise a component of the constructed artificial neural network and extracting the component from the artificial neural network for storage in a component library.

15. The method of claim 14 wherein the component comprises a central pattern generator.

16. The method of claim 12 comprising identifying a causality path of another artificial neural network from an input neuron to an output neuron thereof.

17. The method of claim 12 comprising extracting an unsuccessful substructure from the artificial neural network.

18. The method of claim 12 comprising identifying a substructure of the artificial neural network based on input events received by the substructure.

19. The method of claim 12 comprising identifying a successful substructure of the artificial neural network based on activity of the successful substructure.

20. The method of claim 12 wherein the application is classification, the method further comprising

identifying a first substructure for recognizing a first image and a second substructure for recognizing a second image different from the first image.