DDA4220 Fall 2026
Deep Learning and Applications
- Lectures
- Tue & Thu · 10:30–12:00
- Room
- J_E205
- First lecture
About the course
This course introduces the foundations of deep learning: neural network architectures, training objectives, optimization, regularization, and the engineering practices behind effective models. We explore supervised, self-supervised, unsupervised, and reinforcement learning, with applications in computer vision, natural language processing, and robotics.
You will learn to design, implement, and train models with tools such as PyTorch and TensorFlow, gaining practical experience through hands-on assignments and projects. Advanced topics will be introduced as time permits.
Prerequisites & preparation
Python proficiency is required. The course prerequisites are:
- CSC1001 / CSC1003
- MAT2040 / MAT2041, or equivalent courses
- DDA3020 / CSC4020 / DDA2020 / FTE4560
Prior coursework or projects in computer vision or natural language processing, and familiarity with PyTorch or TensorFlow, are helpful. There are no co-requisites.
Lecture schedule
The course schedule is tentative and subject to change. Lecture materials will be linked below as they become available.
| Lecture | Topics | Materials |
|---|---|---|
| 01 | Course introduction & history | Slides (PDF) |
| 02 | Neural networks: representation and learning | Slides (PDF) |
| 03 | Training neural networks I: optimization | — |
| 04 | Training neural networks II: initialization, normalization and regularization | — |
| 05 | Convolutional neural networks | — |
| 06 | Feature visualization & adversarial examples | — |
| 07 | Transfer learning | — |
| 08 | Object detection | — |
| 09 | Image segmentation | — |
| 10 | Recurrent neural networks & language modeling | — |
| 11 | Sequence-to-sequence models & attention | — |
| 12 | Transformers & vision transformers | — |
| 13 | Midterm review | — |
| Assessment— | Midterm exam · (tentative) | — |
| 14 | Self-supervised learning | — |
| 15 | Autoencoders & variational autoencoders | — |
| 16 | Generative adversarial networks | — |
| 17 | Diffusion models & flow matching | — |
| 18 | Reinforcement learning foundations & value-based methods | — |
| 19 | Policy gradients & actor–critic methods | — |
| 20 | Language foundation models | — |
| 21 | Multimodal & generative foundation models | — |
| 22 | 3D vision | — |
| 23 | Robot learning | — |
| 24 | Advanced topics I | — |
| 25 | Advanced topics II | — |
| 26 | Course review & discussion | — |
Recommended reading
Assessment & policies
- 3 homework assignments
- 45%
- Final projectProject description, milestones, and requirements.
- 25%
- Midterm exam
- 30%
Assignments
- Assignment 1Not yet released
- Assignment 2Not yet released
- Assignment 3Not yet released
Course policies
See L01 for coursework and AI-use guidance. Late-submission policies will be announced.
Teaching team & office hours
Instructor office hours
Tuesday 09:00–10:00 · every other week
First session: . Location to be announced.
Held only by request; please email at least one day in advance.
Lead teaching assistant
ZHANG Wenqian 张文谦
225040483@link.cuhk.edu.cnOffice hours
Tuesday 17:00–18:00
Thursday 14:30–15:30
Zhixin (知新楼) 408
Teaching assistant
WANG Ruixiang 王瑞翔
225040514@link.cuhk.edu.cnOffice hours
Saturday 17:00–18:00
Sunday 17:00–18:00
TxB609
Teaching assistant
PENG Yifan 彭一凡
225040521@link.cuhk.edu.cnOffice hours
Monday 19:00–20:00
Wednesday 19:00–20:00
Zhixin (知新楼) 429
Acknowledgments
This course draws inspiration from Stanford CS231n, Georgia Tech CS 4644/7643 (Danfei Xu), CMU 11-785, and MIT 6.S191. We also borrow a few figures from their teaching materials.