EECE 576L: Advanced Topics in Deep Learning
Overview
EECE 576L examines geometric and probabilistic deep learning and deep reinforcement learning through lectures, student paper presentations, and a research project.
Course Information
| Instructor | Renjie Liao |
|---|---|
| Term | 2026/27 Winter Session, Term 1 (Fall 2026) |
| TA | Yuanpei Gao · yuanpeig@student.ubc.ca |
| Class | Tuesday, 4:00 p.m. – 7:00 p.m. · UBCV · MCLD 2012 (Hector J. MacLeod Building, Floor 2) |
| Piazza | Course forum |
| Office hour | Tuesday, 1:00 p.m. – 2:00 p.m. · KAIS 3047 (Ohm) |
| renjie.liao@ubc.ca |
Use Piazza for questions and Canvas for submissions, grades, deadlines, and changes.
Announcements: First class: Sep. 15. No class Sep. 8 (the classroom may be used for Imagine UBC orientation activities) or Nov. 10 (midterm break).
Detailed overview · Course project · Course policies · Resources
Schedule
Tuesdays, 4:00 p.m. – 7:00 p.m., MCLD 2012. Dates follow the UBC Vancouver calendar; classes do not meet on UBC holidays or during the midterm break. This course does not meet on Sep. 8 or Nov. 10; the first meeting is Sep. 15 and the last is Dec. 1.
The schedule is tentative. Each of the 25 papers has a 30-minute slot, including questions, normally with one presenter. If enrollment exceeds 25, some papers may be jointly presented by two students sharing the same slot. Ten papers (40%) are new selections, marked New. See the presentation guideline.
| Meeting | Date (Tuesday) | Topic and papers | Slides | Suggested readings |
|---|---|---|---|---|
| 1 | Sep. 15 | Introduction; invariance, equivariance, sets and sequences | Lecture 1 Lecture 2 |
Chapter 13, 14 of PML1 book & DL book DeepSets & Transformers & PreNorm & VisionTransformers & SwinTransformers & Chapter 15 of PML1 book |
| 2 | Sep. 22 | Graph neural networks: message passing and graph convolutions | — | Part II of GRL book & Chapter 23 of PML1 book & Chapter 4 of GNN book & GNNs & GGNNs & GAT & Graphormer & GPS & GCNs & ChebyNet & LanczosNet & SignNet & Specformer |
| 3 | Sep. 29 | Group-equivariant deep learning | — | UvAGEDL |
| 4 | Oct. 6 | Paper presentations 1–5 1. PointNet++ 2. Point Transformer V3 3. Mamba 4. Vision Transformers Need Registers 5. G-CNNs |
— | — |
| 5 | Oct. 13 | Paper presentations 6–10 6. Tensor Field Networks 7. SE(3)-Transformers 8. E(n) Equivariant GNNs 9. VGGT New 10. UMA New |
— | — |
| 6 | Oct. 20 | Autoregressive models, LLMs and policy gradients | — | BERT & GPT3 & T5 & Scaling Laws & LoRA |
| 7 | Oct. 27 | Paper presentations 11–15 11. Visual Autoregressive Modeling (VAR) 12. DPO 13. KTO 14. DAPO New 15. Understanding Reasoning from Pretraining to Post-Training New |
— | — |
| 8 | Nov. 3 | Diffusion and flow models | — | Score-based Models & ScoreSDE & DDPM & DDIM & DPM++ Flow Matching & Rectified Flow & Stochastic Interpolants |
| — | Nov. 10 | No class — UBC midterm break | — | — |
| 9 | Nov. 17 | Paper presentations 16–20 16. Does RL Really Incentivize Reasoning Beyond the Base Model? 17. DeepSeek-V3.2 New 18. π0.7: a Steerable Generalist Robotic Foundation Model with Emergent Capabilities New 19. World Action Models are Zero-shot Policies (DreamZero) New 20. GLASS Flows New |
— | — |
| 10 | Nov. 24 | Paper presentations 21–25 21. OT-CFM 22. MDLM 23. LLaDA New 24. MeanFlow New 25. Rectified Flow |
— | — |
| 11 | Dec. 1 | Project presentations | — | — |
FAQ
Can I audit or sit in?
UBC students, staff, and faculty may email the instructor to request permission to audit. Registered students receive priority if classroom space is limited.
Where can I find textbooks, paper lists, and sample projects?
See Course Resources. There is no required textbook.