Gael Varoquaux • 1/22/2020

Survey of machine-learning experimental methods at NeurIPS2019 and ICLR2020

A research survey analyzing the experimental procedures of machine learning researchers who published at NeurIPS 2019 and ICLR 2020. It examines quantitative aspects like hyperparameter optimization methods (mostly manual/grid search), number of baselines, datasets used, and practices for reporting variance (e.g., random seeds). The report highlights trends and potential biases in current ML experimentation, contributing to discussions on reproducibility and benchmarking.

0 comments

#Machine Learning #benchmarking #Reproducibility