Why are transfer learning and pre-trained models commonly used in deep learning? (A) They are only useful for image-related tasks
(B) They enable models to leverage prior knowledge, reducing the need for large labeled datasets
(C) They completely eliminate the need for training on new data
(D) They always outperform models trained from scratch
Why are transfer learning and pre-trained models commonly used in deep learning?
Share
The correct answer is (B) They enable models to leverage prior knowledge, reducing the need for large labeled datasets.
Explanation: Transfer learning and pre-trained models allow a model that has already been trained on a large dataset to apply its learned features to a new, related task. This means that instead of starting from scratch, which often requires substantial labeled data, a model can fine-tune its understanding based on the knowledge acquired from the earlier task. This process significantly reduces the amount of data and time needed for training on the new task.