Generative Programming System and Method Employing Focused Grammars

Abstract

A system and method may employ, focused grammars to facilitate automated generation of computer programs. Such implementation of focused grammars enables a new form of symbolic regression referred to as generative programming or automated programming. The search through the space of possible programs may be guided by a streak search method in accordance with which identified candidate programs that improve on the current streak may be used to create focused grammars for directing an additional localized search. In some embodiments, candidate programs are generated by randomly traversing focused grammars from the starting rule, and subsequently through the grammars, until a complete candidate program parse tree has been constructed. Candidate programs may then be executed, for example, by an evaluator, which may employ an interpreter adapted for use in conjunction with a Stack Manipulation Language or other interpreted language.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS[0001]The present application is a continuation of co-pending U.S. application Ser. No. 11 / 045,928 filed Jan. 28, 2005, and claims the benefit of U.S. provisional application Ser. No. 60 / 544,953, filed Feb. 14, 2004, entitled “GENERATIVE PROGRAMMING THROUGH FOCUSED GRAMMARS: A NEW METHOD OF AUTOMATED PROGRAMMING.” The disclosures of the foregoing applications are hereby incorporated herein by reference in the their entireties.BACKGROUND[0002]1. Field of the Invention[0003]Aspects of the present invention relate generally to advanced symbolic regression programming techniques, and more particularly to a system and method of generative programming which employ focused grammars to facilitate automated generation of computer programs or computerized or electronic solutions.[0004]2. Description of Related Art[0005]Conventional recursive analysis technologies and symbolic regression techniques suffer from various shortcomings. Briefly, traditional st...

AI Technical Summary

Problems solved by technology

Conventional recursive analysis technologies and symbolic regression techniques suffer from various shortcomings.

Conventional methods employing random searches to identify good candidate solutions are deficient to the extent that they tend not address local search areas, i.e., areas within the search space in the vicinity of, or proximate to, other good candidate solutions.

Conversely, genetic programming techniques, while potentially adept at localized searches, are deficient to the extent that the algorithms employed may get caught in local minima to the exclusion of broader, more promising search directions.

Images