Skip to main content
All terms
Foundations

Transfer Learning

Reusing knowledge a model gained on one task to learn a related task faster.

Definition

Transfer learning takes knowledge a model gained on one task or dataset and applies it to a new, related problem, usually by fine-tuning. Starting from a pretrained model rather than from scratch sharply reduces the data and compute a new task needs. It is the standard way to adapt foundation models and is why small teams can build capable applications.