📜 Rules

Our test set is publicly available, which makes it theoretically possible to cheat (e.g. through usage of external data that is not part of the OpenBioLink dataset, hyperparameter searches on the test set, …). While we hope that teams will act ethically and will not attempt to cheat, we will examine the public repository of potential awardees very thoroughly. Any detected misconduct of an awardee team will result in disqualification from the competition.

• We only accept submissions that are fully reproducible (including any Hyperparameter searches).
• The code repositories of all submissions have to be made public.
• We do not allow the usage of external data. Only the usage of data provided by either the python dataloader module or the downloadable resource is allowed.
• It is not allowed to resplit the dataset (i.e. use a larger/smaller validation set) or use the sets other than their intended use (f.e. train a model on the union of training and validation set).
• The intended uses of the different sets are:
  • Training set: Used during the learning process (e.g. to fit parameters) of a model
  • Validation set: Used to tune hyperparameters/model selection
  • Test set: Used only during the final evaluation process
• Persons affiliated with the core organizing team may not participate in challenge.