Resources for Illinois Statisticians attempting to learn deep learning.
- Deep Learning - Yann LeCun
- On the Origin of Deep Learning
- CS 231 @ Stanford
- Loosely follow CS231 assignments. Code from scratch in Python:
- Multiple logistic regression with regularization
- Add a hidden layer, with ReLu activation
- Add mini-batch
- Add dropout
- Add batch-norm
- Add abstraction
- Learn high-level libraries
- Train more complex architectures
- CNN
- RNN
- GAN
- Understand computing resources
- Create local VM with necessary libraries
- Create cloud VM with necessary libraries
- Train on "large" datasets
- CIFAR-10
- ImageNet
- Read current literature