基于自动机主动学习的研究综述  

Research Review of Active Learning Based on Automata

作　　者：夏娟 高俊涛[1] XIA Juan;GAO Juntao(College of Computer&Information Technology,Northeastern Petroleum University,Daqing 163318)

机构地区：[1]东北石油大学计算机与信息技术学院,大庆163318

出　　处：《计算机与数字工程》2024年第12期3536-3540,3642,共6页Computer & Digital Engineering

摘　　要：面向对象分析过程中通常借助有限状态机来表达现实系统的基本行为。自动机学习技术是学习系统行为模型的常用工具。随着自动机学习技术的快速发展和研究的逐步深入,该技术在实践中的应用也日益广泛。大型的软件系统往往包含大量的黑盒组件,其行为模型无法通过手工去进行推断。因此,可以利用主动学习技术完全自动地学习系统或组件行为模型。论文对目前自动机主动学习的学习框架、建模形式和学习算法进行调查、整理和总结归纳。对学习算法的存储结构进行分类,并根据处理反例的不同方式对学习算法进行分析。最后,对自动机主动学习的挑战和未来研究的方向进行了探讨。In the process of object-oriented analysis,the basic behavior of real system is usually expressed by finite state ma-chine.Automata learning technology is a common tool for learning system behavior model.With the rapid development of automatic learning technology and the gradual deepening of research,this technology is increasingly widely used in practice.Large software systems often contain a large number of black box components,and their behavior model cannot be inferred manually.Therefore,ac-tive learning technology can be used to fully automatically learn the system or component behavior model.This paper investigates,arranges and summarizes the curent learning framework,modeling forms and learning algorithms of automatic learning.The storage structure of the learning algorithm is classified,and the learning algorithm is analyzed according to the different ways of dealing with counterexamples.Finally,the challenges of automatic learning and the direction of future research are discussed.

关 键 词：主动学习 黑盒组件 行为模型 学习算法 反例 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]