Course Information

ECS 189G Deep Learning
Date: Jan 8-March 15, 2024
Time: M/W/F 5:10-6:00PM
Venue: ART 204

Instructor Information

Instructor: Jiawei Zhang
Email: jiwzhang[AT]ucdavis.edu
Office: Zoom Link to Join
Office Hours: 6:20-7:20PM (Monday)

TA #1: Zizhong Li (Q&A, Project)
Email: zzoli[AT]ucdavis.edu
Office: Zoom Link to Join
Office Hours: TBD

TA #2: Xinhao Xiang (Q&A, HW, Exam)
Email: xhxiang[AT]ucdavis.edu
Office: Zoom Link to Join
Office Hours: TBD (, Week 2 - 6)

TA #3: Xiao Liu (Q&A, HW, Exam)
Email: xioliu[AT]ucdavis.edu
Office: Zoom Link to Join
Office Hours: TBD (, Week 7 - 10)

Course Description

Do you know how the recent Turing Awards were given to the deep learning field? For what contributions? Deep learning is one of the hottest topics and technology with edge-cutting applications in various areas currently, e.g., search engine, e-commerce, self-driving vehicle, robotics, gaming, and highly like the coming metaverse. What breakthroughs turned the rather mathematical idea into reality? What are the classic milestones that pioneered and shaped the landscape of the deep learning field? What are the state-of-the-art methods just coming out yesterday? Do you know that, although deep learning has been widely used in our real-life products and services now, researchers are even more fascinated by general deep learning issues? The new challenges naturally arise in the junction of deep learning vs class machine learning, deep vs shallow, general-purpose AI vs expert system, even human vs robots, and many more novel contexts. What are the current topics and future agenda? To build the essential foundation as a stepping-stone to deep learning research, this course exposes students to various deep learning concepts, methods, and applications. We will broadly explore the classic as well as more recent research work.

Topics Covered

• basic math, optimization, machine learning background knowledge
• neural network basics, error backpropagation algorithm
• auto-encoder model for data encoding and re-construction
• convolutional neural network (CNN) for computer vision
• recurrent neural network (RNN) for natural language processing
• graph neural network (GNN) for network embeddings
• other advanced topics...

Textbook

Required textbook.

• Deep Learning, 1nd edition, by Ian Goodfellow and Yoshua Bengio and Aaron Courville. MIT Press, 2016.

Recommended textbook.

• Neural Networks and Deep Learning: A Textbook, 1st edition, by Charu Aggarwal. Springer, 2018.
• Neural Networks and Learning Machines, 3rd edition, by Simon S. Haykin. Pearson, 2008.

Prerequisites

Students should come with good programming skills. ECS 060 or ECS 032B or ECS 0036C or equivalent courses are required. If you are not sure whether you have the right background, please contact the instructor. Note: We will not cover programming-specific issues in this course.

Course Format

The course is lecture-based with two examinations (one midterm and one final). There are individual assignments and a group-level programming project. In order to encourage attending classes and participating in discussions, there will be several in-class quizzes for students.

• Lectures and Class Participation: We strongly encourage (and appreciate!) students to attend classes, because effective lectures rely on students' participation to raise questions and contribute in discussions. Although we probably will have a large class, we will strive to maintain interactive class discussions if possible. We will provide lecture notes before class, which will be posted on the Schedule page.
• Questions: We encourage students discussing their questions and problems first with their group peers and classmates. This way, you can get immediate help and also learn to communicate "professionally" with your peers. In any case for more thorough discussion, come to the office hours of TA's and the instructor's. Any announcement will be posted on the Announcement page. Make sure to check it frequently enough to stay informed.
• Assignment: There will be a few written assignments spaced out over the course of the quarter. All the assignments should be done individually by the students. Assignments should be submitted before the class begins on the due dates.
• Exam: There will be a mid-term exam and a final exam held at the mid & end of the quarter.
• Project: There will be a quarter-long project, which involves significant deep learning application programming. The project will be structured with several milestones due in the course of the quarter, leading to a demo and write-up near the end of the quarter.