2025-26

Course Instructor:  Professor Hongyang Li

Teaching Assistants: Yunchao Zhang, PANI Anupam, Hai Zhang

Lecture Time and Venue: Friday 2:00PM - 4:50PM at Graduate House GHP603

Course Description: 

This course aims at introducing the fundamentals in algorithms, data and systems of the autonomous intelligent systems, which often refers to the autonomous driving and robotics applications. As the fast advances in the field of AI, how to utilize the learning-based, data-driven approaches to improve the applications for the better human life, becomes very pivotal. We will address the key challenges in this domain, such as (i) how to formulate a system that is equipped with generalization, intelligence and reliability merits. (ii) How to balance the data distribution between simulation and real-world data. (iii) Is scaling law the only pathway towards high-level AGI. 

We will introduce the concepts, principles and knowhow to build the autonomous intelligent systems. The basic fundamentals would be detailed in the lectures, with tutorials and hands-on training sessions. All the important topics will be covered, such as imitation learning, reinforcement learning, and so on, with a focus on the applications in autonomous driving and robotics. The highlights in this course would consist of several guest lectures from outside renowned speakers from both industry and academia to address the latest advances in this field. The hands-on session is akin to tutorials or hackathons where students learn the recipe of technologies from scratch quickly. These features would be complementary to the main lecture and facilitate the final group presentation.

Course Instructor:  Professor Yi Ma

Teaching Assistants: Tianzhe Chu and Feng Chen

Lecture Time: Mondays 2:00pm - 4:50pm 

Course Description: 

This course aims to provide a rigorous and systematic introduction to the mathematical and computational principles of deep learning. We achieve this by centering the course around a common and fundamental problem behind almost all modern practices of artificial intelligence and machine learning such as image recognition and generation. The problem is how to effectively and efficiently learn a low-dimensional distribution of data in a high-dimensional space and then transform the distribution to a compact and structure representation. Such a representation can be generally referred to as a  memory learned from the sensed data. 

We will start with the most basic and classical cases of PCA, ICA, and Dictionary Learning that assume the distribution has linear and independent structures. To generalize these classical models and solutions to general data distributions, we introduce a universal computational principle for learning low-dimensional distributions: compression. As we will see, data compression provides a unifying view on popular approaches for distribution or representation learning such as Score Matching (for denoising) and coding Rate Reduction. Within this framework, modern deep neural networks, such as ResNet and Transformers, can all be mathematically fully interpreted as (unrolled) optimization algorithms to achieve better compression and representation. To ensure the learned representation to be correct and consistent, we will study the effective Auto-Encoding architecture that consists of both encoding and decoding (say for denoising and diffusion). In order for a learning system to be fully automatic and continuous, we will also study a powerful framework of Closed-Loop Transcription that enables the encoding and decoding networks to self-correct hence self-improve via the ubiquitous mechanism of closed-loop feedback. 

Prerequisites: