INIT ML Lectures Spring 2025

Background

Lecture Format:

• Begin every lecture by reviewing the content of the previous lecture.
• Introduce the concept
  • Begin with a high level conceptualization.
  • Give the theoretical background (if applicable).
    • This will likely include derivations and proofs to, for example, understand where an equation came from rather than just saying, “here use this and plug in the numbers.”
  • Visualization (if applicable).
    • (e.g. 3-D Plots for gradient descent and dimensional partitions of neural networks).

Mar 14, 2025

Lecture 1: Foundations of Machine Learning and Linear Regression

What is my intention with this lecture series?

• I want to take you from knowing nothing about machine learning and give you a road map to understand Transformer models which are the state of the art at the moment.
• I am going on this journey with all of you and as a fellow student. I want to intuitively understand these concepts. I don’t just want to throw an equation and say “here use this.” No, let’s ask the questions of “Why is this relevant? Why are we learning this? How does this fit into what we ultimately want to achieve?”
• My goal is that we can all learn the building blocks of neural networks. If you learn the building blocks then making neural networks will be like making a Lego set. Programmatically, it honestly doesn’t look that different, you will see that we are just strapping small things together and holistically they make up the network. I believe that this first principles approach to learning will solidify your understanding and make you more competent in the rapidly evolving field of Machine Learning.

1. Introduction

• Course overview and learning objectives
• Interest survey
• The landscape of ML and where neural networks fit
• Real-world applications

2. Machine Learning Fundamentals

• What is ML? Supervised vs. unsupervised learning
• The core ML workflow: data → model → prediction → evaluation → refinement