Optimal cessation of training and assessment of accuracy in a given class of neural networks

• US 7,991,717 B1
• Filed: 09/10/2001
• Issued: 08/02/2011
• Est. Priority Date: 09/10/2001
• Status: Expired due to Fees

• Current Assignee: Bush Seismic Technologies, LLC, JohnsonDiversey, Inc
• Original Assignee: JohnsonDiversey, Inc
• Inventors: Bush, Ronald R.
• Primary Examiner(s): HOLMES, MICHAEL B
• Application Number:
  US09/950,190
• Time in Patent Office: 3,613 Days
• Field of Search: 706/21, 706/39, 706/16, 706/25
• US Class Current: 706/21
• CPC Class Codes: G06N 3/08 Learning methods

Abstract

A system, method, and process for configuring iterative, self-correcting algorithms, such as neural networks, so that the weights or characteristics to which the algorithm converge to do not require the use of test or validation sets, and the maximum error in failing to achieve optimal cessation of training can be calculated. In addition, a method for internally validating the correctness, i.e. determining the degree of accuracy of the predictions derived from the system, method, and process of the present invention is disclosed.

Claims

1. A method for configuring an iterative, self-correcting algorithm having an objective function, the method comprising the steps of:

selecting training data;

iterating the algorithm on the selected training data to modify weights in the algorithm; and

relying on characteristics of the objective function to determine when the solution to the algorithm has been reached.

2. The method of claim 1, wherein the iterative, self-correcting, algorithm is a neural network.

3. The method of claim 1, wherein the iterative, self-correcting, algorithm is used to predict producing and non-producing hydrocarbon areas from seismic data.

4. The method of claim 1, wherein the iterative, self-correcting, algorithm is a used to predict producing and non-producing hydrocarbon areas from seismic data in real-time during the acquisition of the said seismic data.

5. The method of claim 1, wherein the iterative, self-correcting, algorithm is embedded in hardware.

6. The method of claim 1, wherein the iterative, self-correcting, algorithm is run on parallel computers.

7. The method of claim 1, wherein the iterative, self-correcting, algorithm is embedded in hardware that is run in parallel.

8. The method of claim 1, wherein the characteristics relied upon include monotonic and asymptotic characteristics.

9. The method of claim 1, wherein the characteristics relied upon include monotonic behavior with an inflection point.

10. The method of claim 1, wherein the iterative, self-correcting, algorithm is configured without using test or validation data.

11. A method for configuring an iterative, self-correcting algorithm having an objective function, the method comprising the steps of:

selecting training data;

using back propagation to train the algorithm by iterating the algorithm on the selected training data to modify weights in the algorithm; and

relying on characteristics of the objective function to determine when the solution to the algorithm has been reached.

12. The method of claim 11, wherein the iterative, self-correcting, algorithm is a neural network.