Chenliang Xu

pic

Spring 2017 - CSC 298/578 Deep Learning and Graphical Models

Information

Time/Place: MW 10:25-11:40 in 301 GAVET

Instructor: Prof. Chenliang Xu
Instructor's Office Hours: MW 15:30-16:30 in 714 CSB

TA: Haofu Liao
TA's Office Hours: MW 12:30-13:30 in 603 CSB

The syllabus is here.


Description

Probabilistic graphical models are a powerful framework that leverages probability theory and graph representation to model complex problems with large collections of inter-related variables. Numerous applications can be found in machine learning, computer vision and beyond. On the other hand, recent developments in neural networks (aka deep learning) have greatly advanced the performance of state-of-the-art in visual recognition and artificial intelligence. This course will cover the essentials (representation, learning and inference) in both probabilistic graphical models and neural networks with an emphasis in analyzing image and video data. Students entering the class should have a pre-existing working knowledge of probability, statistics, linear algebra and programming.


Textbook


Schedule

This is a tentative schedule of the course.
Course materials will be posted on Blackboard.
Green: lectures in DL; Blue: lectures in PGM; Red: other lectures; Yellow: exam/presentation; Grey: no class or cancelled class.

W Date Topic Read Notes
1 1/18 Introduction
2 1/23,25 Foundations: Probability, Graph Theory, Programming KF 2.1, 2.2
3 1/30,2/1 Linear Regression, Gradient Descent, Logistic Regression, Regularization Optional: Andrew Ng's notes; Bishop 1.1, 3.1, 4.3
4 2/6,8 Feedforward, Backpropagation GBC 6; UFLDL Tutorial HW1 Out 2/6, Teaming Due 2/8
5 2/13,15 CNN, Regularization GBC 9, 7
6 2/20,22 Optimization, RNN GBC 8, 10; H. Jaeger 2002 Ch 1, 2
7 2/27,3/1 LSTM, Autoencoders, Applications GBC 14, 11, 12, 15
8 3/6,8 Representation: Bayesian Networks KF 3, 5 HW1 Due 3/6, Proposal Due 3/8, HW2 Out 3/10
9 3/13,15 Spring Recess
10 3/20,22 Representation: Markov Networks KF 4
11 3/27,29 Inference: Variable Elimination, Belief Propagation KF 9, 10
12 4/3,5 Inference: Loopy BP, MAP Graph Cut KF 11, 13 HW3 Out 4/5
13 4/10 Guest Lecture by Prof. Ji Liu HW2 Due 4/10 (ext.)
13 4/12 Inference: Sampling I KF 12
14 4/17, 19 Inference: Sampling II; Learning: Parameter Learning in BNs KF 17
15 4/24, 26 Learning: Parameter Learning in MNs; PGMs in Deep Learning KF 20; GBC 20
16 5/1 Project Presentation (Group 1) All slides due 5/1 8am
16 5/3 Project Presentation (Group 2) HW3 Due 5/3, Project Report Due 5/15 noon (ext.)

Presentation Groups
Group 1: Bot01, Deep South Taco, Doge, Kicking Horse Mountain, Language X, Natty, Synapse, The Adversarial Learners, ZZG.
Group 2: AKI, CAL 9000, Connectome, HAHAHA, Halo, KW, Lorem Ipsum, Naive, Turtlebot.