Spring 26 CSCI 5525 Advanced Machine Learning

Instructor: Zirui “Ray” Liu

Time: Spring 2026, Tues/Thurs 11:15 AM - 12:30 PM

TA: Hao Li

Location: Keller Hall 3-230

Office Hour: Friday 2-3PM. Hybrid. Office Number Keller 1-201, Table #3. Zoom Link

Textbooks

• Pattern Recognition and Machine Learning (PRML) by Chris Bishop
• Deep Learning: Foundations and Concepts by Chris Bishop and Hugh Bishop
• Gaussian Processes for Machine Learning (GPML) by Carl Edward Rasmussen and Christopher K. I. Williams
• Probabilistic Machine Learning: Advanced Topics (PML) by Kevin Murphy

Course Materials

Lecture Slides (password will be sent through Canvas)

Course Description

The course is organized into three parts. In Supervised Learning, we will learn popular ML methods ranging from linear regression, boosting trees, to deep neural networks, and the Transformer architecture that powers modern AI. In Unsupervised Learning, we will learn how machines generate content through autoregressive modeling, Variational Autoencoders, and Diffusion models, which are the engines behind today’s text, image and video generation systems. In Reinforcement Learning, we will learn how AI agents make decisions through policy gradients and Proximal Policy Optimization, the same techniques used to train ChatGPT and create game-playing AI.

Topics Covered

Part I. Supervised Learning

• Machine Learning Basics
• Linear Regression
• Deep Neural Networks
  • Automatic Differentiation
  • Stochastic Optimization
  • Transformer Architectures

Part II. Unsupervised Learning

• Autoregressive Modeling
• Variational Autoencoders
• Diffusion

Part III. Reinforcement Learning

• Reinforcement Learning Basics
• Policy Gradient
• Advantage Estimation & Value Function
• Proximal Policy Optimization

Grading Policy

Assignments can be completed in teams of up to 2 students.

Late Policy

Late submissions will not be accepted.

Academic Integrity

• Each assignment can be carried out by a team of 2 students. We encourage you to find teams early in the semester.
• Use of GenAI tools (e.g., ChatGPT) is allowed, but you must include clear citations for any parts where you use AI.
• Caveat: Research shows that relying entirely on GenAI harms learning. Focus on your own learning rather than simply putting together a submission.

Course Schedule (tentative)

The course schedule may be changed.