← Deep Learning learning pathDeep Learning · Module 8 of 10
Transfer learning and pretrained models
Reuse powerful pretrained models for new tasks.
📖 Key ideas
- Pretrained models learned from huge datasets (ImageNet, web text).
- Transfer learning adapts them to your task with less data.
- Feature extraction freezes the base; fine-tuning updates some layers.
- Hugging Face hosts thousands of pretrained models.
- Saves time, money and data.
💡 Real-life example: Transfer learning is hiring an experienced cook and teaching them your restaurant's menu.
Pretrained modelModel trained on large data before
Fine-tuningFurther training on your data
Freezing layersKeeping weights fixed
Hugging FaceHub of pretrained models