Fig. 1

Challenge cycle overview. For each challenge, participants can form teams of one or more individuals. Challenge teams work together to develop a model (depicted as open box), train their model on training data (purple cylinders) provided by the challenge organizers, containerize their model (closed box with outline), and submit their model to the challenge container repository. Submitted models are run on validation data (green cylinders) on a cloud computing system by the challenge organizers. Once predictions produced by the models are evaluated and scored, results are made available to the challenge teams. Teams can use this information to make improvements to their model and resubmit their optimized model