Multi-module genetic Programming with multiple genetic data representations

• US 20030004904A1
• Filed: 06/28/2001
• Published: 01/02/2003
• Est. Priority Date: 06/28/2001
• Status: Active Grant

• Current Assignee: Valtrus Innovations Limited (f/k/a Dolya Holdco 9 Limited) (Key Patent Innovations Limited)
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Inventors: Kirshenbaum, Evan Randy
• Granted Patent: US 7,424,461 B2
• Time in Patent Office: Days
• US Class Current: 706/13
• CPC Class Codes: G06N 3/126 Evolutionary algorithms, e.g. genetic algorithms or genetic programming

Abstract

A method for genetic programming which is based on multiple genetic data component representations. Multiple genetic data component representations are determined and a solution is evolved using a set of genetic data components represented by the multiple genetic data component representations.

Claims

1. A method for genetic programming, comprising the steps of:

determining more than one genetic data component representation;

evolving a solution using a set of genetic data components represented by the genetic data component representations.

2. The method of claim 1, wherein the step of determining comprises the step of determining a model for the solution such that the model includes a set of modules and the step of evolving comprises the step of parameterizing each module based on the genetic data components.

3. The method of claim 2, wherein the step of evolving the solution comprises the step of generating a child candidate solution by combining a set of genetic data components from each of a set of parent candidate solutions such that a set of genetic data components in the child candidate solution conforms to the genetic data component representations.

4. The method of claim 3, wherein the step of combining comprises the step of combining the genetic data components of the parent candidate solutions which have a common type of genetic data component representation.

5. The method of claim 3, wherein the step of combining comprises the step of combining the genetic data components of the parent candidate solutions which correspond to an analogous module in the child candidate solution.

6. The method of claim 2, wherein the model includes a module that corresponds to a program and a module that corresponds to a set of parameters.

7. The method of claim 6, wherein the genetic data components include a tree and a sequence of numbers and the step of parameterizing includes the steps of using the tree to realize the program and the sequence of numbers to realize the parameters.

8. The method of claim 6, wherein at least one operator of the program references the parameters.

9. The method of claim 2, wherein the model includes a module that corresponds to a program and module that corresponds to a neural network.

10. The method of claim 9, wherein the genetic data components include a first tree and a second tree and the step of parameterizing includes the steps of using the first tree to realize the program and using the second tree to realize the neural network.

11. The method of claim 9, wherein at least one operator of the program queries the neural network.

12. The method of claim 2, wherein the model includes a module that corresponds to a math function and a module that corresponds to a set of parameters.

13. The method of claim 12, wherein the genetic data components include a tree and a sequence of numbers and the step of parameterizing includes the steps of using the tree to realize the math function and using the sequence of numbers to realize the parameters.