This repository offers a hands-on tutorial series on foundational machine learning concepts, designed to accompany the Week 2 lectures of the REU'25 program at AI-EDGE Institute.
Additional notes:
- Environment Setup Guide
- Codebase Structure Overview
- Colab Usage Guide
- Practical Transformer Architectures (Lab slides for module 3,4)
| Module | Title | Subsection | Requires GPU? |
|---|---|---|---|
| 1 | Online Perceptron for Linear Classification | 1.1: A toy example from slide 8 | ❌ CPU-only |
| 1.2: Perceptron on large-margin linearly separable data | ❌ CPU-only | ||
| 1.3: Perceptron on small-margin linearly separable data | ❌ CPU-only | ||
| 1.4: Perceptron on non-linearly separable data | ❌ CPU-only | ||
| 2 | From Taylor Expansions to Gradient Descent | 2.1: Taylor approximation on toy functions | ❌ CPU-only |
| 2.2: Full-batch Gradient Descent | ❌ CPU-only | ||
| 2.3: Compare stochastic vs full-batch Gradient Descent | ❌ CPU-only | ||
| 3 | Transformer for Binary Classification | 3.1: Sequence classification using a Transformer encoder | ✅ CPU / GPU |
| 4 | Transformer for Image Classification | 4.1: Vision Transformer (ViT) on image patches | ✅ GPU |
- ziyueluocs/torch-tutorial for environment setup
- alochaus/taylor-series for sec 2.1
- tintn/vision-transformer-from-scratch for sec 4.1