A new TIER2 paper, published in Wiley’s AI Magazine, addresses the reproducibility challenges and opportunities in machine-learning-based research.

Despite growing concern and appeals from machine learning (ML) researchers, the general community continues to take the matter of reproducibility lightly. Even with a surge in related literature, no comprehensive overview exists of the barriers and drivers that shape reproducibility in the field.

To fill that gap, the paper – co-authored by Harald Semmelrock, Tony Ross-Hellauer, Simone Kopeinik, Dieter Theiler, Armin Haberl, Stefan Thalmann, and Dominik Kowald – categorises the barriers and drivers to four types of ML reproducibility: description, code, data, and experiment, with specific reference to research in both computer science and biomedical fields. 

he authors propose a drivers–barriers matrix that maps common challenges to actionable solutions, including technology-driven solutions, procedural enhancements and enhanced awareness and education – with the latter serving as a crucial foundation that supports and reinforces the other two. Read the full paper here.