It is clear that we need to change train/test splitting in our source codes for preparing final stag submissions.
But I’m not sure what else we are allowed to change in our approaches.
What about:
- model type
- model hyperparameters
- training hyperparameters
- input features
- data sources
Please clarify.