There is little empirical evidence about the testing strategies used by Machine Learning engineers.
In this project, we perform the first fine-grained empirical study on Machine Learning testing practices in the wild. This study aims at shedding some light on the various testing techniques for Machine Learning based software adopted in practice by the research community. This is achieved by identifying the different properties being tested, the testing strategies employed, as well as their implementation throughout the Machine Learning workflow. Our research therefore aims at highlighting the gaps and limitations of existing Machine Learning testing practices.
Équipe du projet
Datasets used for experiments:
Reference :
An Empirical Study of Challenges in Converting Deep Learning Models