Transfer Learning

Last Updated: July 29, 2026 | By Mihail Sebastian | AI Dictionary

The reuse of knowledge a model gained on one task to learn a related task faster, with less data – the strategy behind most modern deep learning.

What is Transfer Learning?

Transfer learning is the reuse of knowledge a model gained on one task to learn a related task, with less data and compute than training from scratch. It works because low-level knowledge generalizes: the edges and textures a vision model learns on everyday photos also describe tumors, and the grammar a language model learns on the web also structures legal contracts.

Nearly every modern deep learning system starts this way rather than from random weights.

How Transfer Learning Works

A source model is trained first on a large dataset; this is what makes a pretrained model. Its learned representations then move to the target task by one of two main routes.

The first is feature extraction: freeze the pretrained layers and train only a new output layer on top, so the old model serves as a fixed encoder. The second is fine-tuning: continue training some or all of the pretrained weights on the target data. Feature extraction suits tiny datasets; fine-tuning wins when there is enough target data to adjust the weights safely.

Transfer Learning vs Fine-Tuning

The practical difference: transfer learning names the general strategy of reusing knowledge across tasks, and fine-tuning is one specific technique for executing it. Every fine-tuned model is an instance of transfer learning, but not all transfer learning involves fine-tuning, since feature extraction reuses knowledge without updating a single original weight.

Transfer learningFine-tuning
What it namesThe broader strategyOne specific technique
What happens to the modelKnowledge is reused; weights may stay frozenWeights update through further training
Starting pointAny knowledge carried over from a source taskAlways a pretrained model
ExampleReusing ImageNet features to read X-raysTraining BERT further on legal contracts

Example of Transfer Learning

A hospital team wants a model that flags skin cancer in dermatology photos, but it has only a few thousand labeled images. The team starts from a network pretrained on ImageNet’s millions of everyday photos.

The early layers, which detect edges, colors, and textures, transfer unchanged; those features describe skin lesions as well as they describe cats and cars. The team retrains the later layers on the dermatology images, and the model learns to separate malignant from benign patterns with a dataset that would be hopeless for training from scratch.

Related AI terms: Fine-tuning · Pretrained Model · Few-Shot Learning · Deep Learning

Did you like the Transfer Learning gist?

Learn about 250+ need-to-know artificial intelligence terms in the AI Dictionary.

Mihail Sebastian — Writes about AI governance, regulation, and the technology behind them. Placeholder bio — replace with a real credential line. About

Read the Governor's Letter

Stay ahead with Governor's Letter, the newsletter delivering expert insights, AI updates, and curated knowledge directly to your inbox.

By subscribing to the Governor's Letter, you consent to receive emails from AI Guv.
We respect your privacy - read our Privacy Policy to learn how we protect your information.

Browse All AI Terms A–Z

Every term in the dictionary, in alphabetical order. Jump to a letter or scroll the full list.

A

B

C

D

E

F

G

H

I

J

K

L

M

N

O

P

Q

R

S

T

U

V

W

X

Y

Z