System and method for constructing cognitive programs

• US 20090216695A1
• Filed: 04/27/2009
• Published: 08/27/2009
• Est. Priority Date: 12/07/2004
• Status: Active Grant

• Current Assignee: Eric Baum
• Original Assignee: Eric Baum
• Inventors: Baum, Eric
• Granted Patent: US 7,853,540 B2
• Time in Patent Office: Days
• US Class Current: 706/13
• CPC Class Codes: G06F 9/30145 Instruction analysis, e.g. decoding, instruction word fields
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

The present invention is directed to a method to search for a solution to a problem in a domain. The method may comprise obtaining a plurality of agents each operable to produce one or more numerical bids and to propose one or more actions and a plurality of nodes each representing a state of the domain; automatically selecting a respective agent and a respective node based on a bids from the plurality of agents; and automatically adding a new node representing a new state which is obtained by applying to the state represented by the selected node an action proposed by the selected agent. The plurality of nodes may each have a depth associated therewith and the respective agent and the respective node may be selected regardless of the depth associated with the selected node.

Claims

1. A method of creating a program to perform a first task having a first goal, the method comprising:

receiving input data pertaining to a number of other tasks each being different from the first task and each having a goal different from the first goal, in which said input data includes information pertaining to a number of examples in which each example represents a particular situation of one of the other tasks;

receiving satisfaction criteria;

automatically obtaining a number of subprograms by learning from the number of examples in the received input data that are determined, in accordance with the satisfaction criteria, operable to satisfactorily perform one of the other tasks;

adding the obtained subprogram or subprograms to a storage library;

creating a number of instructions each operable to invoke one of the subprograms stored in the storage library;

selecting one or more of the created instructions;

adding the selected instruction or instructions to previously obtained instructions to obtain available instructions;

selecting a number of the available instructions so as to create a first instruction set; and

creating said program to perform said first task based on the first instruction set.

2. The method according to claim 1, in which each automatically obtained subprogram is operable to satisfactorily perform one of the other tasks by itself.

3. The method according to claim 2, in which the number of other tasks is at least two, in which the input data includes for each of some of the other tasks information specifying a respective second instruction set which is different from the first instruction set, and in which the respective automatically obtained subprogram which is operative to satisfactorily perform the respective task is constructed from the respective second instruction set.

4. The method according to claim 1, further comprising retrieving from storage a stored program having a number of arguments and a number of associated annotations, in which the associated annotations are utilized in obtaining the number of examples, and in which the creating of said program includes substituting for each argument one of the available instructions.

5. The method according to claim 4 in which at least one of the number of associated annotations is utilized in specifying a second instruction set that is utilized in the automatically obtaining.

6. The method according to claim 4, further comprising selecting a module-construction program from a plurality of stored module-construction programs by utilizing at least one of the number of associated annotations, and in which said learning from examples is performed by running the selected module-construction program.

7. A system for creating a program to perform a first task having a first goal, the system comprising:

means for receiving input data pertaining to a number of other tasks each being different from the first task and each having a goal different from the first goal, in which said input data includes information pertaining to a number of examples in which each example represents a particular situation of one of the other tasks;

means for receiving satisfaction criteria;

means for automatically obtaining a number of subprograms by learning from the number of examples in the received input data that are determined, in accordance with the satisfaction criteria, operable to satisfactorily perform one of the other tasks;

means for adding the obtained subprogram or subprograms to a storage library;

means for creating a number of instructions each operable to invoke one of the subprograms stored in the storage library;

means for selecting one or more of the created instructions;

means for adding the selected instruction or instructions to previously obtained instructions to obtain available instructions;

means for selecting a number of the available instructions so as to create a first instruction set; and

means for creating said program to perform said first task based on the first instruction set.

8. The system according to claim 7, in which each automatically obtained subprogram is operable to satisfactorily perform one of the other tasks by itself.

9. The system according to claim 8, in which the number of other tasks is at least two, in which the input data includes for each of some of the other tasks information specifying a respective second instruction set which is different from the first instruction set, and in which the respective automatically obtained subprogram which is operative to satisfactorily perform the respective task is constructed from the respective second instruction set.

10. The system according to claim 7, further comprising means for retrieving from storage a stored program having a number of arguments and a number of associated annotations, in which the associated annotations are utilized in obtaining the number of examples, and in which the means for creating said program includes means for substituting for each argument one of the available instructions.

11. The system according to claim 10, in which at least one of the number of associated annotations is utilized in specifying a second instruction set that is utilized by the means for automatically obtaining.

12. The system according to claim 10, further comprising means for selecting a module-construction program from a plurality of stored module-construction programs by utilizing at least one of the number of associated annotations, and in which said learning from examples is performed by running the selected module-construction program.

13. A system for creating a program to perform a first task having a first goal, the system comprising:

a receiver operable to receive input data pertaining to a number of other tasks each being different from the first task and each having a goal different from the first goal, in which said input data includes information pertaining to a number of examples in which each example represents a particular situation of one of the other tasks;

a device operable to receive satisfaction criteria;

an obtaining device operable to automatically obtain a number of subprograms by learning from the number of examples in the received input data that are determined, in accordance with the satisfaction criteria, operable to satisfactorily perform one of the other tasks;

a device operable to add the obtained subprogram or subprograms to a storage library;

a device operable to create a number of instructions each operable to invoke one of the subprograms stored in the storage library;

a selecting device operable to enable selection of one or more of the created instructions;

a device operable to add the selected instruction or instructions to previously obtained instructions to obtain available instructions;

a device operable to enable selection of a number of the available instructions so as to create a first instruction set; and

a creating device operable to create said program to perform said first task based on the first instruction set.

14. The system according to claim 13, in which each automatically obtained subprogram is operable to satisfactorily perform one of the other tasks by itself.

15. The system according to claim 14, in which the number of other tasks is at least two, in which the input data includes for each of some of the other tasks information specifying a respective second instruction set which is different from the first instruction set, and in which the respective automatically obtained subprogram which is operative to satisfactorily perform the respective task is constructed from the respective second instruction set.

16. The system according to claim 13, further comprising a device operable to retrieve from storage a stored program having a number of arguments and a number of associated annotations, in which the associated annotations are utilized in obtaining the number of examples, and in which the device operable to create said program includes a device operable to substitute for each argument one of the available instructions.

17. The system according to claim 16, in which at least one of the number of associated annotations is utilized in specifying a second instruction set that is utilized by the obtaining device.

18. The system according to claim 16, further comprising a device operable to select a module-construction program from a plurality of stored module-construction programs by utilizing at least one of the number of associated annotations, and in which said learning from examples is performed by running the selected module-construction program.