Discover how to implement transfer learning using PyTorch, the popular machine learning framework.
- [Jonathan] PyTorch is an increasingly popular deep learning framework and primarily developed by Facebook's AI Research Group. It's popular because of it's Pythonic approach, its flexibility, and because it allows you to run computations immediately. Now, Transfer Learning allows you to use the pretrained parameters of the state-of-the-art deep learning models, as your starting point. This means less time training on expensive GPUs, and you won't require as many training images compared to if you trained a deep learning model from scratch. So you're giving yourself a real head start in most computer-vision related problems, like object classification or detection. Together, PyTorch and Transfer Learning pack quite the punch. I'm Jonathan Fernandes and I work in data science, machine learning, and AI for a consultancy. Join me on LinkedIn Learning, as we look at Transfer Learning for Images in PyTorch.
- What is transfer learning?
- Using autograd
- Creating a fixed feature extractor
- Training an extractor
- Fine-tuning the ConvNet
- Learning rates and differential learning rates