Table of Contents
CSCI 4800: Generative Deep Learning
Course Syllabus
Fall, 2024
Instructor: Shawn McCarthy
Email: shawn.mccarthy@ucdenver.edu
Phone: (303) 900-8613
Class: Microsoft Teams / 330PM (Monday and Wednesday)
Office Hours: 445pm-545pm M/W Teams
Catalog Data: An in-depth examination of generative deep learning, focusing on the design and development of models that can generate new content.
Co-requisites: N/A
Prerequisites:
- CSCI 3412 - Algorithms
Note: Each student must sign and return the attached Prerequisites Agreement form to receive any credit for any assignment or exam. If this form is not returned by the 1st week, the student will be administratively dropped from the course.
Expected Knowledge at the Start of the Course:
- Understanding of data structures and algorithm design
- Familiarity with a high-level programming language, such as Python
Expected Knowledge Gained at the end of the Course:
- Solid understanding of generative deep learning and its applications
- Experience in implementing different types of generative models like VAEs, GANs, and autoregressive models
- Experience with TensorFlow and Keras
- Understanding of the practical considerations in training generative models
ABET Assessment Criteria:
- This is a Computer Science Tech Elective, specifically students design, implement and test Generative Deep Learning in the context of this course.
- (2) Design, implement, and evaluate a computing-based solution to meet a given set of computing requirements in the context of the program’s discipline
Course Objectives: Develop a solid understanding of generative deep learning and its applications. The course will provide hands-on experience with designing, implementing, and training generative models.
Textbook:
- Generative Deep Learning: Teaching Machines to Paint, Write, Compose, and Play
- Generative AI with LangChain: Build large language model (LLM) apps with Python, ChatGPT and other LLMs
Topics: Generative modeling, Deep learning, Variational autoencoders, Generative adversarial networks, Autoregressive models, Normalizing flow models, Energy-based models, Diffusion models, Transformers, Music generation, World models, Multimodal models, Ethics and challenges in generative AI
Course Outline:
| Lecture/Week | Date | Topic | Reading | Assignments |
|---|---|---|---|---|
| 1 | Aug 19th | Introduction to the course and Generative Deep Learning | Preface, Foreword, Objective and Approach, I. Introduction to Generative Deep Learning, 1. Generative Modeling | Read through “Introduction to Generative Deep Learning”. Sign Pre-Req (Algorithms) Sign-up for one of the Student Presentations below (up to 5) |
| 2 | Aug 26th | Deep Learning | 2. Deep Learning | Complete exercises on preparing data, building, compiling, and evaluating the model. Deep Learning (Student Presentation) |
| 3 | Sep 02nd | Variational Autoencoders | 3. Variational Autoencoders | Implement a basic Variational Autoencoder on Fashion-MNIST Dataset. Variational Autoencoders (Student Presentation) |
| 4 | Sep 09th | Generative Adversarial Networks Lab Week | Lab1: LLM (Building Transformer-Based Natural Language Processing Applications) 4. Generative Adversarial Networks | Implement a simple Generative Adversarial Network. |
| 5 | Sep 16th | Autoregressive Models | 5. Autoregressive Models | Implement an autoregressive model and generate a sequence. Generative AI (Student Presentation) Autoregressive Models (Student Presentation) |
| 6 | Sep 23rd | Normalizing Flow Models | 6. Normalizing Flow Models | Experiment with a simple normalizing flow. Normalizing Flow Models (Student Presentation) |
| 7 | Sep 30th | Energy-Based Models | 7. Energy-Based Models | Create a simple energy-based model. Energy Based Models (Student Presentation) |
| 8 | Oct 07th | Diffusion Models Lab Week | Lab 2: Nvidia Diffusion 8. Diffusion Models | Implement a diffusion model. |
| 9 | Oct 14th | Transformers for Generative Models | 9. Transformers for Generative Models | Implement a Transformer model for sequence generation. Diffusion Models (Student Presentation) Transformers for Generative Models (Student Presentation) |
| 10 | Oct 21st | Music Generation | 10. Music Generation | Develop a simple model for music generation. Music Generation (Student Presentation) |
| 11 | Oct 28th | World Models | 11. World Models | Implement a simple world model. |
| 12 | Nov 04th | Multimodal Models Lab Week | Lab3: RAG (Building RAG agents with LLMs) 12. Multimodal Models | Implement a model that combines different types of data. |
| 13 | Nov 11th | Ethics and Challenges in Generative AI | 13. Ethics and Challenges in Generative AI | Discussion and reflection on the ethical considerations and challenges in generative AI. |
| 14 | Nov 18th | Project Presentations | Present final projects. | |
| 15 | Nov 25th | Thanksgiving Break | ||
| 16 | Dec 02nd | Course Wrap-up | Course reflection. |
Grading Policy:
- Homework: 65%
- Projects: 35%
Notes: UCD Code of Honor as in the catalog: UCD Honor Code
Projects (you can work individual or as a team of up to 5)
- Applications based on Part III
- Langchain based applications
Mental Health Resources: CU Denver faculty and staff understand the stress and pressure of college life. Students experiencing symptoms of anxiety, depression, substance use, loneliness or other issues affecting their mental well-being, have access to campus support services such as the Student and Community Counseling Center, the Wellness Center and the Office of Case Management. Students also have access to the You@CUDenver on-line well-being platform available 24/7. More information about mental health education and resources can be found at Lynx Central and CU Denver’s Health & Wellness page. Students in imminent crisis can contact Colorado Crisis Services for immediate assistance 24/7 or walk-in to the counseling center during regular business hours.
Courses in this program
Meet your instructor
Shawn McCarthyFAQs
Are there prerequisites?
- CSCI 3412 - Algorithms
How often do the courses run?
- Every Fall