🧬 CS-844 — Generative Deep Models

Department of Computer Engineering
College of Electrical and Mechanical Engineering (CEME)
National University of Sciences and Technology (NUST)

---

📋 Table of Contents

• 📌 Course Overview
• ℹ️ Course Information
• 📅 Course Schedule
• 🎯 Course Objectives
• 📚 Weekly Topics
• 🎓 Course Learning Outcomes
• 📊 Assessment & Grading
• 📖 Course Materials
• 🏛️ Institution

---

📌 Course Overview

This course provides a comprehensive introduction to modern deep generative models, assuming basic familiarity with probability, calculus, and Python programming .
Starting from the foundations of generative modeling, it develops the intuition, mathematical underpinnings, and applications of key generative model families including autoregressive models, latent variable models, GANs, energy-based and score-based models, diffusion models, and large language models (LLMs) .

---

ℹ️ Course Information

As per the CS-844 Generative Deep Models course outline, Department of Computer Engineering, CEME — NUST .

Field	Details
📘 Course Number	CS-844
🏫 Course Title	Generative Deep Models
🎓 Credits	(as specified in official outline)
👩‍🏫 Instructor(s)-in-charge	As per department allocation
🏛️ Department	Computer Engineering, CEME — NUST
🎯 Course Type	Core / Elective (as in outline)
📚 Prerequisites	Basic probability, calculus, and Python
🎓 Degree & Semester	Postgraduate (PG)
📅 Academic Year	As mentioned in the official course outline

You can fill in instructor name, credits, semester, and year exactly as they appear in your final approved outline.

---

📅 Course Schedule

The course outline specifies the standard NUST schedule structure for lectures, discussions, and self-study .

Component	Details
🕐 Lectures	Weekly lecture-based delivery (3 hrs/week typical PG pattern)
🔬 Lab	As specified in the official outline (lab / no-lab)
💬 Discussion	Discussion/tutorial hours as allocated by the instructor
📖 Outside Study	Independent study and project work as recommended
🏢 Office Hours	Defined by the instructor each semester

You can refine these bullets to match the exact hours once confirmed.

---

🎯 Course Objectives

The main objectives of CS-844 as stated in the outline are :

• Understand probabilistic foundations of deep generative models .
• Develop a deep understanding of key model families including autoregressive models, VAEs, GANs, and diffusion models .
• Apply generative models to complex high-dimensional data (e.g., images, text, audio) .
• Explore real-world applications of deep generative models in modern AI systems .

These objectives emphasize both rigorous theory and practical implementation on real datasets.

---

📚 Weekly Topics

The following structure mirrors the “Week / Lecture Title / Topics” table in your course outline .

Week	Lecture Title	Topics
1	Introduction to the Course	Overview of the course, motivation for generative modeling, applications and examples .
2	Generative Models	Background on generative modeling, density estimation, joint vs conditional modeling .
3	Autoregressive Models	From classical generative models to deep generative models, intro to PyTorch; autoregressive factorization .
4	Maximum Likelihood Learning	Learning generative models, likelihood, gradient-based optimization, basics of neural networks .
5–7	Latent Variable Models	Variational Autoencoders (VAEs): vanilla VAEs, semi-supervised VAEs, disentangled VAEs .
8	Normalizing Flows	Invertible models, change-of-variables formula, flow architectures .
9–10	GANs & Energy-Based Models	Generative Adversarial Networks, BiGANs, adversarial training, energy-based formulations .
11–13	Diffusion & Score-Based Models	Score-based diffusion models, discrete latent variable models, diffusion for discrete data .
14–15	Large Language & Multimodal Models	Introduction to LLMs, multimodal models such as CLIP, modern applications of generative AI .

---

🎓 Course Learning Outcomes

Your outline maps Course Learning Outcomes (CLOs) to Program Learning Outcomes (PLOs), focusing on complex AI-based software engineering problems .

CLOs and Mapped PLOs

CLO	Description	Mapped PLO
CLO-1	Apply advanced knowledge of machine learning, probability, and engineering fundamentals to explain and solve complex AI-based software engineering problems using generative models .	PLO-1: Engineering Knowledge
CLO-2	Identify, formulate, and analyze complex machine learning problems using generative modeling, probabilistic reasoning, and first principles of engineering sciences .	PLO-2: Problem Analysis
CLO-3	Design conceptual solutions for complex machine learning problems by selecting appropriate generative modeling techniques while considering system requirements and constraints .	PLO-3: Design/Development of Solutions
CLO-4	Conduct analytical investigation of generative models via literature review, experiments, and interpretation of results to reach valid conclusions .	PLO-4: Investigation
CLO-5	Recognize the need for continuous learning and critical thinking to adapt generative modeling knowledge to emerging technologies and evolving software engineering challenges .	PLO-11: Lifelong Learning

Use these CLOs explicitly in your assignment and project descriptions in the repo.

---

📊 Assessment & Grading

The outline defines the standard NUST assessment structure including exams, homework, lab/design reports, quizzes, and participation .

Assessment Components

Component	Description
🧪 Exam(s)	Midterm / final exam(s) covering theoretical and applied aspects of generative models .
📝 Homework	Written and coding assignments on topics like VAEs, GANs, diffusion, etc. .
🧪 Lab / Design Reports	Reports linked to implementation projects or experiments with generative models .
📄 Design Report	A conceptual or empirical project report focused on a selected generative model or application .
🗒️ Quizzes	Short quizzes to regularly assess understanding of core concepts .
👥 Class Participation	Active participation in discussions, presentations, and code reviews .

💡 Exact percentage weights (e.g., 10% quizzes, 25% assignments, etc.) should be copied from the finalized assessment table in your official outline or course handout.

---

📖 Course Materials

The outline specifies a section for textbooks, references, and additional material .

Suggested Structure

Primary Textbook(s)
- Title, Author(s), Edition, Publisher, Year

Reference Books
- Title, Author(s), Publisher, Year
- ...

Additional Material
- Research papers on VAEs, GANs, diffusion models, LLMs
- Official docs for PyTorch / JAX / HuggingFace
- Selected blog posts and tutorials (for intuition)