Retraining and Fine-Tuning Strategies

Background: The effectiveness of deep learning models for emerging disease forecasting relies heavily on using pre-trained models and either retraining or fine-tuning them as new, pathogen-specific data becomes available.

Question / Future Work: While the current study used a one-time training approach, operational deployment in a real-time setting could involve either retraining the model weekly as new COVID-19 data emerges or, more commonly, performing fine-tuning on the pre-trained model using the new, pathogen-specific data.