Special Topics in AI: Deep Learning

CS839, Fall 2020
Department of Computer Sciences
University of Wisconsin–Madison

Tentative Schedule (Subject to Change)

Week Date Topic Reading materials Assignments
1 Thursday, September 3 Course overview and introduction [Slides]
Sign up for paper presentations and scribes (link to Google sheet)
2 Tuesday, September 8 Neural Architecture Design I (lecture) [Slides] Goodfellow-Bengio-Courville Chapter 6
2 Thursday, September 10 Neural Architecture Design II [Slides][Notes] ImageNet Classification with Deep Convolutional Neural Networks
Deep Residual Learning for Image Recognition
Densely Connected Convolutional Networks (to present)
Mask R-CNN
3 Tuesday, September 15 Neural Architecture Design III [Slides][Notes] Neural Architecture Search with Reinforcement Learning
Efficient Neural Architecture Search via Parameter Sharing (to present)
3 Thursday, September 17 Trustworthy Deep Learning I (lecture) [Slides] Goodfellow-Bengio-Courville Chapter 7.5, 7.13
Team registration (link to Google sheet)
4 Tuesday, September 22 Trustworthy Deep Learning II [Slides][Notes] Out-of-distribution Reliability
Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images
On Calibration of Modern Neural Networks
Enhancing The Reliability of Out-of-distribution Image Detection in Neural Networks
A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attacks
Generalized ODIN: Detecting Out-of-distribution Image without Learning from Out-of-distribution Data
Informative Outlier Matters: Robustifying Out-of-distribution with Outlier Mining (to present)
Energy-based Out-of-distribution Detection
4 Thursday, September 24 Trustworthy Deep Learning III [Slides][Notes] Adversarial Robustness
Towards Deep Learning Models Resistant to Adversarial Attacks
Intriguing properties of neural networks
Robust Physical-World Attacks on Deep Learning Models
Unlabeled Data Improves Adversarial Robustness
Defense Against Adversarial Images using Web-Scale Nearest-Neighbor Search
Understanding and Mitigating the Tradeoff Between Robustness and Accuracy (to present)
5 Tuesday, September 29 Trustworthy Deep Learning IV [Slides][Notes] Fairness / Group Robustness
An investigation of why overparameterization exacerbates spurious correlations (to present)
Invariant Risk Minimization
Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalization
Model Patching: Closing the Subgroup Performance Gap with Data Augmentation
5 Thursday, October 1 No class (use the time to discuss the project proposal with team)
Project proposal submission deadline on October 4 Midnight (download proposal latex template here)
6 Tuesday, October 6 Meetings with Instructor to Review Project Proposals
Signup sheet for 15-min slot
8 Thursday, October 8 Interpretable Deep Learning I (Lecture) [Slides] Guest lecture by TA: Yiyou Sun
9 Tuesday, October 13 Interpretable Deep Learning II [Slides][Notes] Attribution Methods
Learning Deep Features for Discriminative Localization (CAM)
Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization (to present)
Network Dissection: Quantifying Interpretability of Deep Visual Representations
Axiomatic Attribution for Deep Networks
9 Thursday, October 15 Interpretable Deep Learning III [Slides][Notes] Rethinking interpretability
Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead (to present)
Interpretation of Neural Networks is Fragile
The (Un)Reliability of Saliency Methods
10 Tuesday, October 20 Deep Learning Generalization and Theory I (Lecture) [Slides] Guest speaker: Yuandong Tian (Facebook AI Research)
10 Thursday, October22 Deep Learning Generalization and Theory II [Slides][Notes] Exploring Generalization in Deep Learning
Towards Understanding the Role of Over-Parametrization in Generalization of Neural Networks (to present)
On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima
Understanding deep learning requires rethinking generalization
Visualizing the Loss Landscape of Neural Nets
11 Tuesday, October 27 Deep Learning Generalization and Theory III [Slides][Notes] One ticket to win them all: generalizing lottery ticket initializations across datasets and optimizers
The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks (to present)

6 Thursday, October 29 Learning with Less Supervision I (Lecture) [Slides]
7 Tuesday, Novermber 3 Learning with Less Supervision II [Slides][Notes] Weakly supervised learning
Big Transfer (BiT): General Visual Representation Learning (to present)
Exploring the Limits of Weakly Supervised Pretraining
Learning Visual Features from Large Weakly Supervised Data
Revisiting Unreasonable Effectiveness of Data in Deep Learning Era
Snorkel: Rapid Training Data Creation with Weak Supervision
7 Thursday, November 5 Learning with Less Supervision III [Slides][Notes] Unsupervised learning
SCAN: Learning to Classify Images without Labels
A Simple Framework for Contrastive Learning of Visual Representations (to present)
Self-Supervised Learning of Pretext-Invariant Representations
Rethinking Pre-training and Self-training
8 Tuesday, November 10 Learning with Less Supervision IV [Slides][Notes] Data Augmentation (some blogs I wrote)
Automating the Art of Data Augmentation: Overview
Automating the Art of Data Augmentation: Practice
Automating the Art of Data Augmentation: Theory
Automating the Art of Data Augmentation: New Direction
Unsupervised Data Augmentation for Consistency Training (to present)
11 Thursday, November 12 Continual / Lifelong Learning I (Lecture) [Slides]
12 Tuesday, November 17 Continual / Lifelong Learning II [Slides][Notes] Overcoming Catastrophic Forgetting with Unlabeled Data in the Wild (to present)
iCaRL - Incremental Classifier and Representation Learning
LwF-Learning without Forgetting
Overcoming catastrophic forgetting in neural networks
12 Thursday, November 19 Continual / Lifelong Learning III [Slides][Notes] SupSup - Supermasks in Superposition (to present)
PackNet: Adding Multiple Tasks to a Single Network by Iterative Pruning
Progressive Neural Networks
13 Tuesday, November 24 Deep Generative Model I (Lecture) [Notes on GAN][Notes on style transfer GAN] Goodfellow-Bengio-Courville Chapter 20
13 Thursday, November 26 No class (Thanksgiving)
14 Tuesday, December 1 Deep Generative Model II [Slides][Notes] Generative Adversarial Networks
InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets
Improved Techniques for Training GANs
Stacked Generative Adversarial Networks
Large Scale GAN Training for High Fidelity Natural Image Synthesis
Wasserstein GAN (to present)
14 Tuesday, December 3 Deep Generative Model III Style Transfer
Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
A Neural Algorithm of Artistic Style (to present)
A Style-Based Generator Architecture for Generative Adversarial Networks
15 Tuesday, December 8 Final project presentation (Team 1-9)
15 Tuesday, December 10 Final project presentation (Team 10-17)
16 Thursday, December 17 Final project written report due (by end of the day)