Perspective of Software Engineering Researchers on Machine Learning Practices Regarding Research, Review, and Education

The paper investigates how software engineering researchers approach machine learning in their work, reviewing research, review, and education practices. It finds that while many researchers follow data collection, model training, and evaluation routines, only a minority adopt recommended practices such as hyperparameter tuning. Common challenges include data handling, evaluating non‑functional properties, and integrating human expertise, and education often relies on hands‑on activities alongside traditional methods.