Discovering Boundary Values of Feature-based Machine Learning Classifiers through Exploratory Datamorphic Testing

Abstract

Testing has been widely recognised as di cult for AI applications. is paper proposes a set of testing strategies for testing machine learning applications in the framework of the datamorphism testing methodology. In these strategies, testing aims at exploring the data space of a classi cation or clustering application to discover the boundaries between classes that the machine learning application de nes.is enables the tester to understand precisely the behaviour and function of the so ware under test. In the paper, three variants of exploratory strategies are presented with the algorithms implemented in the automated datamorphic testing tool Morphy. e correctness of these algorithms are formally proved. eir capability and cost of discovering borders between classes are evaluated via a set of controlled experiments with manually designed subjects and a set of case studies with real machine learning models.