Transfer learning
Starting from a model already trained on a large, related dataset and adapting it to a new, usually smaller, task rather than training from random weights. It works because early layers tend to learn general representations that transfer across tasks, so only the later layers need much retraining.
Why exams ask this
Tested as "why start from a pretrained model instead of training from scratch." The exam wants the answer framed as reused representations saving both data and compute, and flags a distractor that claims transfer learning always outperforms training from scratch regardless of how different the new task is.
Relevant to
Related concepts
Resources
No resources linked to this concept yet.