What we work on
Data converters, training configurations, model evaluation and the website. A good change is small, linked to an issue and verifiable in a reproducible way. Remember: a faster model does not make an experiment more convincing if the split or budget changed at the same time.
Time and a first task
Plan for about four hours a week. A first 30-minute step: propose a test on synthetic IDs that checks one patient never appears in both training and evaluation. Agree where the test belongs in an issue first; do not download medical data.
Good first issues. If none fits, describe your idea in a new issue.
What to include in a pull request
- the minimal change and why it is needed;
- the command to check it and any environment specifics;
- which data versions, metrics or interfaces it touches.
Large architectural changes are discussed first. Technical details are in the project documentation. An honest “I did not run this check” beats an unverified tick.
Computer science students: what the project offers you and topics for term papers and theses.