In the context of transfer learning, what is a potential risk when adapting a pre-trained model to a new task?

  • Difficulty in fine-tuning hyperparameters.
  • Loss of all knowledge from the source task.
  • Overfitting to the source domain.
  • Unavoidable decrease in model performance.
A potential risk in transfer learning is the loss of all knowledge from the source task. When adapting a pre-trained model to a new task, if not done carefully, the model might forget what it learned during the source task, which can hinder its performance on the new task.
Add your answer
Loading...

Leave a comment

Your email address will not be published. Required fields are marked *