Lectures are held at WTS A30 (Watson Center) from 9:00am to 11:15m on Monday (starting on Jan 13, 2020). The second lecture is from 9:00am to 11:15am on Friday (Jan 17, 2020).
Instructors: Dr. Martin Renqiang Min and Prof. Mark Gerstein.
A detailed list of topics to be covered is here.
Technical sections: Friday 9:00am - 10:00am DL 120 (Dunham Laboratory).
Teaching Fellow: Yitan Wang.
Textbook: There is no required textbook for this course, but there are several freely available online textbooks for your reference.
1. Deep Learning.
2. Dive into Deep Learning.
3. Neural Networks and Deep Learning.
4. Pattern Recognition and Machine Learning [pdf].
A closely related course focusing on computer vision was offered at Stanford (cs231n).
| Event Type | Date | Description | Course Materials |
|---|---|---|---|
| Lecture 1 | Monday Jan 13 |
Course Introduction Introduction to neural networks, backpropagation, and deep learning Course logistics |
[slides]
Required Reading: [backpropagation] Deep Learning Review Paper [Nature] Optional Reading: [backprop notes] [Efficient BackProp] related: [1], [2], [3] |
| Lecture 2 | Friday Jan 17 |
Supervised Deep Learning Activation functions, Convoutional Neural Networks Image Classification, AlexNet, VGG, ResNet pretraining, naive transfer learning network visualization, deep dream, style transfer |
[slides]
Required Reading: Deep Learning Review Paper [Nature] AlexNet Optional Reading: [python/numpy tutorial] [image classification notes] [linear classification notes] VGGNet, GoogLeNet, ResNet |
| Technical Section | Friday Jan 24 |
Python/numpy/Deep Learning Hardware/Software |
[Numpy notebook] [PyTorch notebook] Stanford cs231n 2017 YouTube Lecture 8 |
| Lecture 3 | Monday Jan 27 |
Optimization, Regularization, and Robustness Optimization and regularization methods Adversarial examples and robust optimization Attack and defense methods |
[slides]
Optional Reading: SGD by Leon Bottou [cs231n optimization note 1] [cs231n optimization note 2] [cs231n optimization note 3] Robust optimization against adversarial |
| A1 Posted | Wednesday Jan 29 |
Assignment 1 perceptron, backpropagation, optimization programming: experimenting with activation functions, different layers, loss functions, gradient vanishing, and optimization methods |
[A1 Written Part] |
| Technical Section | Friday Jan 31 |
PyTorch and CNN Filter Visualization PyTorch tutorials on Autograd Training a simple CNN and a classifier CNN filter visualization DeepDream and Style Transfer |
Neural Network in PyTorch Classifier in PyTorch Stanford 2017 cs231n YouTube Lecture 12 |
| Lecture 4 | Monday Feb 3 |
Recurrent Neural Networks LSTM, GRU |
[slides]
Optional Reading: DL book RNN chapter min-char-rnn |
| Project Topics | Thursday Feb 6 |
Some possible project topics | [Some Topics] |
| Technical Section | Friday Feb 7 |
Adversarial Examples |
[Stanford 2017 cs231n YouTube Lecture 16] |
| A1 Coding Part | Saturday Feb 8 |
A1 Coding Posted experimenting with activation functions, shallow and deep architectures, dropout, gradient vanishing, and optimization methods |
[A1 Coding Part] [A1 Starter Code] |
| Lecture 5 | Monday Feb 10 |
Deep Autoencoder RBM, Autoencoder |
[slides]
Required Reading: Dimensionality Reduction Application Denoising Autoencoder Optional Reading: Ladder Network tips/tricks: [1], [2] |
| Proposal Title due | Monday Feb 10 |
Project abstract and team formation due | [Stanford 2017 cs231n proposal description] |
| Lecture 6 | Monday Feb 17 |
Deep Encoder-Decoder Networks and Applications Word embedding Machine translation Image and video captioning |
[slides] Required Reading: Seq2Seq Optional Reading: [Stanford 2017 cs231n YouTube Lecture 11] Word2Vec |
| Lecture 7 | Monday Feb 24 |
Attention Mechanisms and Applications Neural Machine Translation Question Answering Transformer and BERT |
[slides] Required Reading: NMT with Attention Optional Reading: Transformer BERT Interactive QA |
| A2 Posted | Thursday Feb 27 |
Assignment #2 posted Understand exploding and vanishing gradient of vanilla RNN, understand RBM and autoencoder PyTorch with DNN, CNN, vanilla RNN, LSTM/GRU |
[Assignment #2] [Starter Code (163M)] |
| Lecture 8 | Monday Mar 2 |
Deep Probabilistic Generative Models: Variational Inference and Variational Autoencoder (Reweighted) Wake-Sleep Algo (Neural) VI, VAE, PixelCNN |
[slides] Required Reading: VAE Optional Reading: Neural VI |
| Lecture 9 (Zoom) | Monday Mar 23 |
Deep Generative Models: Generative Adversarial Networks GAN, Conditional GAN CycleGAN, Domain Adaptation Wasserstein distance, WGAN Video Generation, Text2Video |
[slides] Required Reading: GAN Optional Reading: Wasserstein GAN Text2Video CycleGAN [Stanford 2017 cs231n YouTube Lecture 13] |
| A2 Due | Friday Mar 27 |
Assignment #2 due Understand exploding and vanishing gradient of vanilla RNN, understand RBM and autoencoder PyTorch with DNN, CNN, vanilla RNN, LSTM/GRU |
[Assignment #2] |
| A3 Posted | Saturday Mar 28 |
Assignment #3 posted Understand issues of VAE and GAN Train VAE or GAN on MNIST |
[Assignment #3] |
| Lecture 10 (Zoom) | Monday Mar 30 |
Deep Reinforcement Learning Deep Q-Learning Policy Gradient, Actor-Critic |
[Part1]
[Part2]
Required Reading: AlphaGO Optional Reading: Stanford cs231n 2017 Lecture 14 AlphaZero ICML 2017 DRL Tutorial |
| Lecture 11 | Monday Apr 6 |
Biomedical Application Case Study I |
[slides]
|
| Lecture 12 | Monday Apr 13 |
Biomedical Application Case Study II |
[Part1]
[Part2]
|
| Guest Lecture by Renjie Liao | Friday Apr 17 |
Graph Neural Networks | Renjie Liao |
| Lecture 13 | Monday Apr 20 |
Biomedical Application Case Study III |
[slides]
|
| Guest Lecture by Dr. Asim Kadav | Friday Apr 24 |
Video Understanding | [Slides] |
| A3 Due | Monday Apr 27 |
Assignment #3 due Understand issues of VAE and GAN Train VAE or GAN on MNIST |
[Assignment #3] |
| Paper Summary Due | Thursday Apr 30 |
Paper Summary due 10 Required Papers |
|
| Final Project Paper Due | Thursday 7 May |
Project Paper due | |
| Presentation Video Due | 7 May | Video No More Than 5 Minutes |