Research on Deep Learning Operator Testing of Intelligent Software Based on Fuzz Testing Technology
摘要
This paper introduces the mainstream deep learning framework and deep learning operators, studies the accuracy problems and cause analysis of deep learning operators, and proposes a guided fuzz testing method based on differential testing to test deep learning operators by studying the current mainstream intelligent software testing methods differential testing and fuzz testing, and provides the scheme design and testing operation process.