AI Lecture series (Special topics of Applied AI @ HKU)

Joint work with Dr Jing WANG (University of Greenwich).

This course focuses on generative AI, covering both its theory and implementation.

A key feature of our approach is how we explain the theory. We start by treating a real, open-weight model as a "black box" to build practical and lasting knowledge. Then, by examining its source code, we gradually turn that black box into a "white box."

• 1. Real-World Problems and Task Solving with Artificial Intelligence [lecture note] [slides] [Google Colab material]
• 2. Machine Learning Models and Parametric Functions [lecture note] [slides] [Google Colab material]
• 3. Functions Represented by Neural Networks [lecture note] [slides] [Google Colab material]
• 4. Licenses and Openness [lecture note] [slides] [Google Colab material]
• 5. Pipeline and Tokenization [lecture note] [slides] [Google Colab material]
• 6. Probabilistic Language Model and Token Generation [lecture note] [slides] [Google Colab material]
• 7. Embedding [lecture note] [slides] [Google Colab material]
• 8. Evaluating LLM [lecture note] [slides]
• 9. Evaluating LLM Difference [lecture note] [slides] [Google Colab material]
• 11. Image Generation [lecture note] [slides] [Google Colab material]
• 12. VAE [lecture note] [slides] [Google Colab material]
• 13. Reverse Diffusion [lecture note] [slides] [Google Colab material]
• 14. Diffusion Strategy [lecture note] [slides] [Google Colab material]
• 15. Convolutional neural networks in image generation [lecture note] [slides] [Google Colab material]
• 16. Text encoder and prompt weighting [lecture note] [slides] [Google Colab material]
• 17. Fundamentals of AI tuning [lecture note] [slides] [Google Colab material]
• 18. Parameter efficient finetuning and LoRA [lecture note] [slides] [Google Colab material]
• 19. Learning theory and LoRA training [lecture note] [slides] [Google Colab material]
• 20. Supervision and Reward [lecture note] [slides] [Google Colab material]