Transfer Learning

Executive Definition: Transfer Learning — A machine learning technique where an AI model trained on one task is repurposed as the starting point for a different, related task.

Transfer learning leverages knowledge acquired from a primary task to accelerate learning on a secondary task. By starting with a pre-trained model rather than training from scratch, developers significantly reduce compute costs and data requirements.

Frequently Asked Questions

Is fine-tuning the same as transfer learning?

Fine-tuning is a specific, highly common type of transfer learning. The broader category also includes feature extraction and domain adaptation.

Why does transfer learning work?

Neural networks learn general features (like edges in images or grammar in text) in early layers. Only the final, task-specific layers need to be retrained for the new application.

When does transfer learning fail?

It struggles when the source and target domains are fundamentally dissimilar. For example, a model trained on everyday English will not transfer well to interpreting complex medical imaging.