Glossary Entry

Transfer Learning

Reusing knowledge learned on one task or dataset to help solve a different but related task more efficiently.

Training Reuse

Seed source: Google ML Glossary

Transfer learning starts with a model that has already learned useful structure elsewhere and then repurposes that knowledge for a new problem. This usually reduces the amount of task-specific data and compute you need.

That is the core idea behind several workflows on the blog, from adapting image models to downstream tasks through to fine-tuning language models instead of training from scratch.