Transfer Learning is the technique of reusing a model trained on one task as the starting point for a different but related task, transferring its learned knowledge rather than starting over.
What it means in plain English
Learning from scratch every time is wasteful. Transfer learning takes the general knowledge a model gained on a big task and applies it to a new, often smaller task. Because the model already understands a lot, it needs far less data and time to become good at the new job.
This is one of the ideas that made modern AI practical. It means individuals and smaller organisations can build capable models without the enormous data and computing power needed to train from zero.
A simple example
A model trained to recognise everyday objects in millions of photos can be adapted, with only a few hundred labelled examples, to spot a specific type of manufacturing defect — because it already understands shapes, edges, and textures.
Why it matters
Transfer learning is why powerful AI is accessible beyond the largest tech companies. Fine-tuning a pre-trained model is the most common form of it, and it underpins a huge amount of practical AI development.
Related terms
- Fine-Tuning — the most common way transfer learning is applied.
- Pre-training — produces the model whose knowledge gets transferred.
- Machine Learning — the broader field transfer learning belongs to.
Frequently asked questions
What is transfer learning?
It is reusing a model trained on one task as the starting point for a related task, so you benefit from what it already learned instead of training from scratch.
Why is transfer learning useful?
It saves time and data and often improves results, since the pre-trained model already understands general features — fine-tuning a foundation model is a form of transfer learning.
