自动程序设计──模拟进化的途径  

AUTOMATIC PROGRAMMING:A SIMULATED EVOLUTIONARY APPROACH

作　　者：吴少岩[1] 陈火旺[1] 

机构地区：[1]国防科学技术大学计算机科学系

出　　处：《计算机学报》1997年第2期97-104,共8页Chinese Journal of Computers

基　　金：国家863高技术基金

摘　　要：本文基于一种新颖的进化计算模型，提出一个自动程序设计框架．通过引入程序概括的概念，将给定任务的程序设计转换为寻优问题．程序进化器以形式文法作为遗传表示，它是独立于具体任务的进化模拟机．当程序概括的结果经联编器翻译后，程序进化器能够自动地构造出完成给定任务的目标程序．本文用符号回归任务的自动程序设计检验这一框架的可行性．Based on a novel evolutionary computation model, an automatic programming framework is proposed. By means of the concept of program generalization, programming for a particular task is converted into an optimization problem.Program evolver, which uses a formal grammar as its genetic representation, is amachine of simulated evolution independent of particular tasks. Once results of program generalization are translated by a binder, program evolver can construct thegoal program of a particular task automatically. In this paper, automatic programming for symbolic regression tasks is used to demonstrate feasibility of the framework.

关 键 词：自动程序设计 进化计算 搜索 优化 

分 类 号：TP311.51[自动化与计算机技术—计算机软件与理论]