Testing and Quality Validation for AI Software–Perspectives, Issues, and Practices

This is the authors' abstract. We don't add key points for this paper.

With the fast growth of artificial intelligence and big data computing technologies, more and more software service systems have been developed using diverse machine learning models and technologies to make business and intelligent decisions based on their multimedia input to achieve intelligent features, such as image recognition, recommendation, decision making, prediction, etc. Nevertheless, there are increasing quality problems resulting in erroneous testing costs in enterprises and businesses. Existing work seldom discusses how to perform testing and quality validation for AI software. This paper focuses on quality validation for AI software function features. The paper provides our understanding of AI software testing for new features and requirements. In addition, current AI software testing categories are presented and different testing approaches are discussed. Moreover, test quality assessment and criteria analysis are illustrated. Furthermore, a practical study on quality validation for an image recognition system is performed through a metamorphic testing method. Study results show the feasibility and effectiveness of the approach.