Course Overview
Meeting Information
Time: Monday & Wednesday, 3:30-5:00 PM
Location: Towne Building, Room 337
Start Date: August 27th, 2025
Instructor: Jiatao Gu
Prerequisites
- Graduate-level machine learning
- Deep learning fundamentals
- Linear algebra and probability theory
- Programming experience (Python/PyTorch preferred)
Course Description
This graduate seminar course will provide an in-depth exploration of deep generative modeling, focusing on both foundational principles and recent advances. We will critically examine key model paradigms, including variational autoencoders, generative adversarial networks, normalizing flows, diffusion models, and autoregressive models, emphasizing their theoretical underpinnings, design trade-offs, and emerging trends in both learning and applications. Through student-led discussions and paper presentations, participants will engage with cutting-edge research and open challenges in the field.
Learning Objectives
- Understand the theoretical foundations of modern generative models
- Analyze and implement state-of-the-art generative architectures
- Evaluate recent research papers in the field
- Develop skills in presenting and critiquing research
- Gain experience with practical applications of generative models
Schedule
Loading Course Schedule...
The detailed weekly schedule is loading. This will include topics, readings, presenters, and important dates.
Requirements & Grading
Paper Presentations
Weight: 50%
Students will present 2 research papers throughout the semester, providing critical analysis and leading class discussion.
Class Participation
Weight: 20%
Active participation in discussions, asking insightful questions, and engaging with presented material.
Final Project
Weight: 30%
Original research project involving implementation, experimentation, or theoretical analysis. Includes proposal, progress report, final paper, and presentation.
Assignment Timeline
- Week 3: Paper presentation assignments
- Week 5: Project proposal due
- Week 10: Project progress report
- Week 15: Final project presentations
- Finals Week: Final project reports due
Resources
Textbooks & References
- Deep Learning by Ian Goodfellow, Yoshua Bengio, and Aaron Courville
- Pattern Recognition and Machine Learning by Christopher Bishop
- Recent conference papers from NeurIPS, ICML, ICLR, CVPR
Tools & Frameworks
- Python programming environment
- PyTorch or TensorFlow
- Jupyter notebooks, Google Colab
- Git for version control
- Computing resources will be provided as needed
Useful Links
- OpenAI Diffusion Models - Research on diffusion-based generation
- Denoising Diffusion PyTorch - Clean implementation of diffusion models
- Lilian Weng's Blog - Excellent explanations of generative models
- OpenAI Sora (2024) - Text-to-video generation model
- FLUX.1 Model (2024) - Advanced text-to-image generation
- Stable Diffusion Repository - Open source text-to-image generation
- Runway Gen-3 Alpha (2024) - Video generation research
- Google Genie 3 - New Frontier for World Models
- OpenAI o1 (2024) - Advanced reasoning language model
- Llama 3.1 (2024) - Open source large language model
- Google Gemini 2.0 (2024) - Multimodal large language model
- Trending LLMs (2024-2025) - Latest language models
- Trending Generative Models (2024-2025) - Latest pre-trained models
Contact Information
Instructor
Name: Jiatao Gu
Email: jgu32@cis.upenn.edu
Office Hours: [To be announced]
Office: AGH 423
Course Communication
Course announcements and materials will be distributed via:
- Course website (this page)
- Email announcements
For questions, please email the instructor or visit during office hours.