When and Why Test Generators for Deep Learning Produce Invalid Inputs: an Empirical Study
This work addresses the reliability of automated testing for deep learning systems, which is crucial for developers and testers, though it is incremental as it builds on existing validation methods.
The study investigated the extent to which Test Input Generators (TIGs) produce valid inputs for testing deep learning systems, finding that 84% of generated inputs are valid according to automated validators, but these validators only achieve 78% accuracy in aligning with human assessments.
Testing Deep Learning (DL) based systems inherently requires large and representative test sets to evaluate whether DL systems generalise beyond their training datasets. Diverse Test Input Generators (TIGs) have been proposed to produce artificial inputs that expose issues of the DL systems by triggering misbehaviours. Unfortunately, such generated inputs may be invalid, i.e., not recognisable as part of the input domain, thus providing an unreliable quality assessment. Automated validators can ease the burden of manually checking the validity of inputs for human testers, although input validity is a concept difficult to formalise and, thus, automate. In this paper, we investigate to what extent TIGs can generate valid inputs, according to both automated and human validators. We conduct a large empirical study, involving 2 different automated validators, 220 human assessors, 5 different TIGs and 3 classification tasks. Our results show that 84% artificially generated inputs are valid, according to automated validators, but their expected label is not always preserved. Automated validators reach a good consensus with humans (78% accuracy), but still have limitations when dealing with feature-rich datasets.