📊 Data Science Week 1 of 14 MSc · S2 ⏱ ~50 min

Week 1: Vision Transformers, Diffusion Models & Graph Neural Networks

Advanced deep learning at MSc level: Vision Transformers, Diffusion Models, Graph Neural Networks, and cutting-edge optimization techniques.

University of America

DS505 — Lecture 1 · MSc S2

🎬 CC Licensed Lecture

0:00 / —:—— 📺 MIT OpenCourseWare (CC BY-NC-SA)

🎯 Learning Objectives

Implement a Vision Transformer (ViT) from scratch
Train a diffusion model for image generation
Build and train a Graph Neural Network
Implement advanced optimizers: Adam, LAMB, Lion

Topics Covered This Lecture

Vision Transformers (ViT, SWIN)

Diffusion Models: DDPM & DDIM

Graph Neural Networks

Advanced Optimization Algorithms

📖 Lecture Overview

This first lecture establishes the foundational framework for Deep Learning & Neural Networks. By the end of this session, you will have the conceptual grounding and practical starting point needed for the rest of the course.

        Why this matters
        Advanced deep learning at MSc level: Vision Transformers, Diffusion Models, Graph Neural Networks, and cutting-edge optimization techniques. This lecture sets up everything that follows — make sure you understand the core concepts before proceeding to Week 2.
      

Key Concepts

The lecture introduces the four main pillars of this course: Vision Transformers (ViT, SWIN), Diffusion Models: DDPM & DDIM, Graph Neural Networks, Advanced Optimization Algorithms. Each will be explored in depth over the 14-week curriculum, with hands-on projects reinforcing theory at every stage.

# Quick Start: verify your environment is ready for DS505
import sys
print(f"Python {sys.version}")

# Check key libraries are installed
try:
    import numpy, pandas, matplotlib
    print("✅ Core libraries ready")
except ImportError as e:
    print(f"❌ Missing: {e} — run: pip install numpy pandas matplotlib")

This Week's Focus

Focus on mastering: Vision Transformers (ViT, SWIN) and Diffusion Models: DDPM & DDIM. These are the prerequisites for everything in Week 2. The concepts build on each other — do not skip the practice exercises.

📋 Project 1 of 3 50% of Final Grade

DS505 Project 1: Custom Diffusion Model

Implement a denoising diffusion probabilistic model (DDPM) from scratch in PyTorch. Train on a small image dataset and generate novel samples. Analyze the FID score.

Full DDPM implementation in PyTorch
Training procedure with noise schedule analysis
Generated image samples (20+)
FID score computation and quality analysis

50%

3 Projects

20%

Midterm Exam

30%

Final Exam

📝 Sample Exam Questions

These represent the style and difficulty of questions you'll see on the midterm and final. Start thinking about them now.

Conceptual Short Answer

Explain the forward and reverse processes in a DDPM. What is being maximized during training?

Analysis Short Answer

How does a Vision Transformer differ architecturally from a ResNet-50?

Applied Code / Proof

What is a message passing algorithm in a Graph Neural Network? Write the mathematical update rule.