1、Can extra training dataset or pretrained models be used in the competition?
2、Can we train an updated model using some other training data between validation phase and testing phase and submit the result inferenced by the updated model as the final result?
I think these should be written into the rules of the competition.
Thanks!
1. Yes, the participants can use more training data.
2. In machine learning, the models should not be tuned specifically for test data. So in theory, it is not allowed to tune the model after the validation phase. We will set a deadline right after the validation phase for submitting the trained models, and then the models are not allowed to be changed. Detailed rules and important dates will be updated later.