基于不变量的机器学习算法分析  

Analysis of machine learning algorithms based on invariants

作　　者：宁一璇 NING Yixuan(College of Information,Zhejiang Sci-Tech University,Hangzhou 311121,China)

机构地区：[1]浙江理工大学信息学院,杭州311121

出　　处：《智能计算机与应用》2021年第5期6-12,共7页Intelligent Computer and Applications

基　　金：国家自然科学基金(111229A4A15153)。

摘　　要：随着机器学习算法在人工智能各领域的广泛应用,人们开始关注机器学习算法的质量分析。由于机器学习算法中缺少测试语言,对学习算法进行质量分析是很困难的,基于此,本文提出了基于不变量的机器学习算法分析方法,对5种机器学习算法进行不同参数下的不变量生成,得到不变量集合。通过动态筛选机制和函数调用图,对不变量集合进行筛选和提取,得到机器学习算法的关键变量。实验结果表明,当机器学习算法的质量随着参数的改变越来越好时,关键变量也呈现逐渐减小的趋势,从而可以对机器学习算法进行质量分析。With the widespread application of machine learning algorithms in various fields of artificial intelligence,people have begun to pay attention to the quality analysis of machine learning algorithms. Due to the lack of test predictions in the machine learning algorithm,it is very difficult to analyze the quality of the learning algorithm. Based on this,this paper proposes a machine learning algorithm analysis method based on invariants,which generates invariants under different parameters for five machine learning algorithms,Get the invariant set. Through the dynamic screening mechanism and function call graph,the invariant set is screened and extracted to obtain the key variables of the machine learning algorithm. The experimental results show that when the quality of the machine learning algorithm gets better and better as the parameters change,the key variables also show a gradual decrease trend,so that the quality of the machine learning algorithm can be analyzed.

关 键 词：机器学习算法 程序不变量 关键变量提取 程序分析 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]