An Empirical Investigation of Replicability in Machine Learning Research

This study presents an empirical analysis of the replicability of machine learning research. We conducted a systematic effort to re implement a corpus of 255 papers published across three decades, documenting key methodological features and assessing the success of replication. The research focuses on quantifying the relationship between specific paper characteristics and the ability of independent researchers to reproduce reported results. Statistical analyses are employed to identify factors that significantly influence the replicability of machine learning studies, providing insights into improving the reliability and validity of scientific findings in the field.