CSCI 8945: Advanced Representation Learning (Fall 2019)

Instructor Sheng Li
Time and Location of Lecture TR: 11:00 am - 12:15 pm Boyd GSRC 306
W: 11:15 pm - 12:05 pm Boyd GSRC 306
Office Hours and Location Wednesday: 1:00 pm - 2:00 pm or by an email appointment.
Boyd GSRC 804

Course Description

This course presents a rigorous overview of advanced representation learning algorithms in machine learning, from the traditional subspace learning models to the recent deep representation learning models. Applications in the fields of computer vision, data mining, and natural language processing will be covered.

Please refer to the course syllabus for more details.

Required Prerequisite

CSCI 6380 (Data Mining), CSCI 6550 (Artificial Intelligence) or Permission of Department


Section Portion Description
Paper Review 20% Write comprehensive paper reviews for assigned papers
Paper Presentation 20% Present assigned papers and lead discussion
Team Project 45% Project proposal (10%); Progress review (5%); Final presentation (15%) and report (15%)
In-class Participation 15%

Grade Conversion Table:

Letter Grade A A- B+ B B- C+ C C- D+ D D- F
Range [93,100] [90,92] [87,89] [83,86] [80,82] [77,79] [73,76] [70,72] [67,69] [63,66] [60,62] [0,59]

Academic Honesty

We will strictly follow UGA’s Academic Honesty Policy. Dishonest behavior will not be tolerated and may result into failing the course. Please contact the instructor if you have any concerns regarding this issue.

Paper Review and Presentation

Note: Please sign up for paper presentations (Deadline: 11:59pm, August 16, 2019)

Paper Review: Students will be required to review 10 papers over the semester and submit the review to eLC by 11:59PM of the due date shown in class schedule below. The reviewed paper should be chosen from the papers that will be presented in the following classes (Refer to class schedule for more details). The reviewed papers should not contain the papers you’re going to present in the class. The review should summarize the main idea and contributions of the paper, describe the major experimental results, and discuss the strengths and weaknesses of the paper. The students are also encouraged to check the follow-up works on the topic of the assigned paper (e.g., search the latest papers that cite the assigned paper), summarize the state-of-the-art methods and results, and discuss possible future research directions.

Paper Presentation: Each student or group (2 students) will be required to present two research papers over the semester. Each presenter should prepare slides for a 25 minutes talk on the paper. Slides for the talk must be emailed to the instructor by midnight (12:00am) before the class. The talk should clearly address the following points: (1) motivation and problem statement; (2) related work; (3) methodology; (4) experiments; (4) conclusions; and (5) at least 3 questions for discussion. The presenter will need to lead another 10 minutes discussion during or after the talk. The presenter should prepare discussion questions that lead to a deeper analysis of the paper’s content, strengths, weaknesses, and future works.

Class Schedule (Tentative)

Week Date Topic Papers Presenters Notes
1 08/14 (W) Course Overview N/A Instructor N/A
08/15 (R) An Overview of Machine Learning (I) N/A Instructor
2 08/20 (T) An Overview of Machine Learning (II) N/A Instructor
08/21 (W) Subspace Learning N/A Instructor Review-1 Due for papers: 08/22, 08/27
08/22 (R) Subspace Learning (1) Two-dimensional PCA: a new approach to appearance-based face representation and recognition. IEEE Trans. PAMI, 2004.
(2) Two-Dimensional Linear Discriminant Analysis, NIPS 2004
3 08/27 (T) Subspace Learning (1) Subclass Discriminant Analysis. IEEE Trans. PAMI, 2006
(2) General Tensor Discriminant Analysis and Gabor Features for Gait Recognition. IEEE Trans. PAMI, 2007
08/28 (W) Manifold Learning N/A Instructor Review-2 Due for papers: 08/29, 09/03, 09/04
08/29 (R) Manifold Learning (1) Graph Embedding and Extensions: A General Framework for Dimensionality Reduction. IEEE Trans. PAMI, 2006
(2) Clustering and Dimensionality Reduction on Riemannian Manifolds, CVPR 2008
4 09/03 (T) Manifold Learning (1) Adaptive Manifold Learning, IEEE Trans. PAMI, 2011
(2) Robust Multiple Manifolds Structure Learning, ICML 2012
09/04 (W) Deep Auto-Encoders (1) Extracting and composing robust features with denoising autoencoders, ICML 2008 Instructor and Student Review-3 Due for papers: 09/05, 09/10, 09/11
09/05 (R) Deep Auto-Encoders (1) Contractive Auto-Encoders: Explicit Invariance During Feature Extraction, ICML 2011
(2) Semi-Supervised Recursive Autoencoders for Predicting Sentiment Distributions, EMNLP 2011
5 09/10 (T) Deep Auto-Encoders (1) k-Sparse Autoencoders, ICLR 2014
(2) Semantic Autoencoder for Zero-Shot Learning, CVPR 2017
09/11 (W) Deep Auto-encoders (1) Auto-Encoding Variational Bayes, ICLR 2014
(2) Variational Autoencoders for Collaborative Filtering, WWW 2018
Students Review-4 Due for papers: 09/12, 09/17, 09/18
09/12 (R) Convolutional Networks (1) ImageNet Classification with Deep Convolutional Neural Networks, NIPS 2012
(2) Very deep convolutional networks for large-scale image recognition, ICLR 2015
6 09/17 (T) Convolutional Networks (1) Going Deeper with Convolutions, CVPR 2015
(2) Deep Residual Learning for Image Recognition, CVPR 2016
09/18 (W) Convolutional Networks/td> (1) Densely Connected Convolutional Networks, CVPR 2017
(2) Res2Net: A New Multi-scale Backbone Architecture, IEEE TPAMI, 2019
Students Review-5 Due for papers: 09/19, 09/24
09/19 (R) Convolutional Networks (1) Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks, NIPS 2015
(2) Mask R-CNN, ICCV 2017
7 09/24 (T) Convolutional Networks (1) Learning Spatiotemporal Features with 3D Convolutional Networks, ICCV 2015
(2) Xnor-net: Imagenet classification using binary convolutional neural networks, ECCV 2016
09/25 (W) Project Proposal Presentation (I) N/A Students
09/26 (R) Project Proposal Presentation (II) N/A Students
8 10/01 (T) Project Proposal Presentation (III) N/A Students
10/02 (W) Recurrent Networks Instructor Review-6 Due for papers: 10/03, 10/08, 10/09
10/03 (R) Recurrent Networks (1) Bidirectional recurrent neural networks, IEEE Trans. Signal Processing, 1997 Students
9 10/08 (T) Recurrent Networks (1) Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation, EMNLP 2014
(2) Hierarchical Attention Networks for Document Classification, ACL 2016
10/09 (W) Recurrent Networks (1) Relational recurrent neural networks, NeurIPS 2018 Students Review-7 Due for papers: 10/10, 10/15
10/10 (R) Recurrent Networks (1) Long-Term Recurrent Convolutional Networks for Visual Recognition and Description, IEEE Trans. PAMI, 2016
(2) Attention Is All You Need, NIPS 2017
10 10/15 (T) Recurrent Networks (1) End-To-End Memory Networks, NIPS 2015
(2) Understanding and Controlling Memory in Recurrent Neural Networks, ICML 2019
10/16 (W) Adversarial Networks Instructor Review-8 Due for papers: 10/17, 10/22, 10/23, 10/24
10/17 (R) Adversarial Networks (1) Generative Adversarial Nets, NIPS 2014
(2) Image-to-Image Translation with Conditional Adversarial Networks, CVPR 2017
11 10/22 (T) Adversarial Networks (1) Unsupervised representation learning with deep convolutional generative adversarial networks, ICLR 2016
(2) Wasserstein Generative Adversarial Networks, ICML 2017
10/23 (W) Adversarial Networks (1) Distributional Smoothing with Virtual Adversarial Training, ICLR 2016
(2) Adversarial Training Methods for Semi-Supervised Text Classification, ICLR 2017
10/24 (R) Adversarial Networks (1) Defense-gan: Protecting classifiers against adversarial attacks using generative models, ICLR 2018
(2) Towards Diverse and Natural Image Descriptions via a Conditional GAN, ICCV 2017
12 10/29 (T) Project Progress Update Students
10/30 (W) Graph Neural Networks Instructor Review-9 Due for papers: 10/31, 11/05, 11/06
10/31 (R) Graph Neural Networks (1) Convolutional neural networks on graphs with fast localized spectral filtering, NIPS 2016
(2) Semi-Supervised Classification with Graph Convolutional Networks, ICLR 2017
13 11/05 (T) Graph Neural Networks (1) Dual Graph Convolutional Networks for Graph-Based Semi-Supervised Classification, WWW 2018
(2) Inductive Representation Learning on Large Graphs, NIPS 2017
11/06 (W) Graph Neural Networks (1) Cluster-GCN: An Efficient Algorithm for Training Deep and Large Graph Convolutional Networks, KDD 2019
(2) Hierarchical Graph Representation Learning with Differentiable Pooling, NeurIPS 2018
Students Review-10 Due for papers: 11/07, 11/12, 11/14, 11/19
11/07 (R) Graph Neural Networks (1) GraphRNN: Generating Realistic Graphs with Deep Auto-regressive Models, ICML 2018
(2) Simplifying Graph Convolutional Networks, ICML 2019
14 11/12 (T) Graph Neural Networks (1) Explainability Methods for Graph Convolutional Neural Networks, CVPR 2019
(2) Efficient Relative Attribute Learning Using Graph Neural Networks, ECCV 2018
11/13 (W) AutoML (1) AMC: AutoML for Model Compression and Acceleration on Mobile Devices, ECCV 2018 Students
11/14 (R) AutoML
15 11/19 (T) Course Project--Final Presentation (I) Students
11/20 (W) Course Project--Final Presentation (II) Students
11/21 (R) Course Project--Final Presentation (III) Students
16 11/26 (T) Course Project--Final Presentation (IV) Students
11/27 (W) No Class; Thanksgiving Break
11/28 (R) No Class; Thanksgiving Break
17 12/03 (T) Invited Talk on Reinforcement Learning Prof. Prashant Doshi