Result caching for improving statistical efficiency of composite simulation models

Patent metadata

Field	Value
Publication number	US-9824167-B2
Application number	US-201514743888-A
Country	US
Kind code	B2
Filing date	Jun 18, 2015
Priority date	Jun 18, 2015
Publication date	Nov 21, 2017
Grant date	Nov 21, 2017

Abstract

Embodiments of the present invention provide a method for result caching to improve the statistical efficiency of a composite model comprising a first component model and at least one other component model. The method comprises developing at least one metamodel for at least one component model of the composite model, computing a re-use factor that maximizes an asymptotic efficiency measure based on the metamodel, determining a number of executions to run for the composite model, and determining a number of executions to run for the first component model based on the re-use factor and the number of executions to run for the composite model. The number of executions to run for the first component model is fewer than the number of executions to run for the composite model. Each output generated from each execution of the first component model is cached and provided as input to another component model.

First claim

What is claimed is:

1. A method for result caching to improve the statistical efficiency of a composite model comprising a first component model and at least one other component model, the method comprising:
developing at least one metamodel for at least one component model of the composite model, wherein the at least one metamodel is based on prior observations of the at least one component model, and wherein the at least one metamodel approximates a statistical property of the at least one component model;
computing a re-use factor that maximizes an asymptotic efficiency measure based in part on the at least one metadmodel;
determining a number of executions to run for the composite model;
determining a number of executions to run for the first component model based on the re-use factor and the number of executions to run for the composite model, wherein the number of executions to run for the first component model is fewer than the number of executions to run for the composite model; and
for each execution of the first component model, caching an output generated from the execution, wherein the cached output is provided as input during one or more executions of the at least one other component model.

2. The method of claim 1 , wherein the statistical property of the at least one component model represents expected execution costs of the at least one component model.

3. The method of claim 2 , wherein the expected execution costs incorporate costs of data transformations of each output generated from each execution of the first component model.

4. The method of claim 1 , wherein the statistical property of the at least one component model represents statistical covariances between outputs generated from executions of the first component model and the at least one other component model.

5. The method of claim 1 , wherein one or more cached outputs generated from executions of the first component model are cycled, such that each cached output is provided as input during at least one execution of the at least one other component model.

6. The method of claim 1 , wherein the number of executions to run for the composite model estimates an expected value of a particular performance measure of interest for a given computational budget within a specified confidence interval with a specified probability.

7. The method of claim 1 , wherein the particular performance measure of interest is one of the following: execution costs, profits, and reliability.

8. A system comprising a computer processor, a computer-readable hardware storage medium, and program code embodied with the computer-readable hardware storage medium for execution by the computer processor to implement a method for result caching to improve the statistical efficiency of a composite model comprising a first component model and at least one other component model, the method comprising:
developing at least one metamodel for at least one component model of the composite model, wherein the at least one metamodel is based on prior observations of the at least one component model, and wherein the at least one metamodel approximates a statistical property of the at least one component model;
computing a re-use factor that maximizes an asymptotic efficiency measure based in part on the at least one metadmodel;
determining a number of executions to run for the composite model;
determining a number of executions to run for the first component model based on the re-use factor and the number of executions to run for the composite model, wherein the number of executions to run for the first component model is fewer than the number of executions to run for the composite model; and
for each execution of the first component model, caching an output generated from the execution, wherein the cached output is provided as input during one or more executions of the at least one other component model.

9. The system of claim 8 , wherein the statistical property of the at least one component model represents expected execution costs of the at least one component model.

10. The system of claim 9 , wherein the expected execution costs incorporate costs of data transformations of each output generated from each execution of the first component model.

11. The system of claim 8 , wherein the statistical property of the at least one component model represents statistical covariances between outputs generated from executions of the first component model and the at least one other component model.

12. The system of claim 8 , wherein one or more cached outputs generated from executions of the first component model are cycled, such that each cached output is provided as input during at least one execution of the at least one other component model.

13. The system of claim 8 , wherein the number of executions to run for the composite model estimates an expected value of a particular performance measure of interest for a given computational budget within a specified confidence interval with a specified probability.

14. The system of claim 8 , wherein the particular performance measure of interest is one of the following: execution costs, profits, and reliability.

15. A computer program product comprising a computer-readable hardware storage medium having program code embodied therewith, the program code being executable by a computer to implement a method for result caching to improve the statistical efficiency of a composite model comprising a first component model and at least one other component model, the method comprising:
developing at least one metamodel for at least one component model of the composite model, wherein the at least one metamodel is based on prior observations of the at least one component model, and wherein the at least one metamodel approximates a statistical property of the at least one component model;
computing a re-use factor that maximizes an asymptotic efficiency measure based in part on the at least one metadmodel;
determining a number of executions to run for the composite model;
determining a number of executions to run for the first component model based on the re-use factor and the number of executions to run for the composite model, wherein the number of executions to run for the first component model is fewer than the number of executions to run for the composite model; and
for each execution of the first component model, caching an output generated from the execution, wherein the cached output is provided as input during one or more executions of the at least one other component model.

16. The computer program product of claim 15 , wherein the statistical property of the at least one component model represents expected execution costs of the at least one component model.

17. The computer program product of claim 16 , wherein the expected execution costs incorporate costs of data transformations of each output generated from each execution of the first component model.

18. The computer program product of claim 15 , wherein the statistical property of the at least one component model represents statistical covariances between outputs generated from executions of the first component model and the at least one other component model.

19. The computer program product of claim 15 , wherein one or more cached outputs generated from executions of the first component model are cycled, such that each cached output is provided as input during at least one execution of the at least one other component model.

20. The computer program product of claim 15 , wherein the

Assignees

• IBM

Inventors

• Haas Peter J

Classifications

• G06F8/76

  Adapting program code to run in a different environment; Porting · CPC title

• G06F2113/26

  Composites · CPC title

• G06F30/20 Primary

  Design optimisation, verification or simulation (optimisation, verification or simulation of circuit designs G06F30/30) · CPC title

• G06F8/75

  Structural analysis for program understanding · CPC title

• G06F2111/08

  Probabilistic or stochastic CAD · CPC title