David McAllester, April 2017

Lecture slides for 21 topics in deep learning with pointers into Deep Learning by Goodfellow, Bengio and Courville, as well as pointers to other relevant material.

- Multi-Layer Perceptrons (MLPs) and Stochastic Gradient Descent (SGD)
- Feed-Forward Computation Graphs, Backpropagation, and the Educational Framework (EDF)
- Minibatching in EDF
- Variants of SGD
- An SGD Progress Theorem
- Architecture and Universality
- Convolutional Neural Networks (CNNs)
- Some Linear Systems and Wavelet Theory
- Second Order Optimization Methods
- Vanishing Gradients, Xavier Initialization, Batch Normalization and Highway Architectures (Resnets, LSTMs and GRUs)
- Regularization
- Some Generalization Theory
- Interpreting Deep Networks
- Sequence to Sequence Models and Attention
- Deep Reinforcement Learning
- AlphaGo
- Deep Graphical Models
- Unsupervised and Predictive Learning
- Information Theory and Distribution Modeling
- Variational Autoencoders
- A Rate-Distortion Case Study
- Generative Adversarial Networks (GANs)