节点文献

基于机器学习的类型推理方法综述

A review of type inference methods based on machine learning

  • 推荐 CAJ下载
  • PDF下载
  • 不支持迅雷等下载工具,请取消加速工具后下载。

【作者】 袁梦霆谢婧

【Author】 YUAN Meng-ting;XIE Jing;School of Computer Science, Wuhan University;

【机构】 武汉大学计算机学院

【摘要】 类型推理是一种轻量级的形式化方法,通过对程序变量和语句的类型这些关键信息进行推理,可以更好地理解程序行为.传统的类型推理方法依赖于语法规则与类型推演规则,然而,随着软件技术的发展,在动态语言等新的软件应用场景中,传统的类型推理方法在缺乏运行时信息的时候无法在静态对类型进行推理.针对这些问题,近年来出现了很多基于机器学习的类型推理的方法.基于机器学习的方法,可以利用已有的动态类型信息,对新程序的类型进行静态的类型推理.文章系统地总结了各种基于机器学习进行类型推理的方法,总结其特点和存在的问题,并讨论了未来可能的研究方向.

【Abstract】 Type inference is a kind of lightweight formalized method, which can better understand the behavior of program by deducing the key information of program variables and statement types. The traditional type inference method relies on grammar rules and type inference rules. However, with the development of software technology, in the new software application scenarios such as dynamic language, the traditional type inference method is unable to infer the type statically when it lacks the runtime information. To solve these problems, many methods of type inference based on machine learning have emerged in recent years. The method based on machine learning can use the existing dynamic type information to make static type inference for the new program. In this paper, we systematically summarize various methods of type inference based on machine learning, their characteristics and existing problems, and discuss possible future research directions.

【关键词】 类型推理机器学习程序分析
【Key words】 type inferencemachine learningprogram analysis
【基金】 国家自然科学基金资助项目(61872272,61640221)
  • 【文献出处】 广州大学学报(自然科学版) ,Journal of Guangzhou University(Natural Science Edition) , 编辑部邮箱 ,2019年03期
  • 【分类号】TP181;TP311.1
  • 【下载频次】249
节点文献中: 

本文链接的文献网络图示:

本文的引文网络