Skip to content
Free public beta until December 4, 2026: every course is open, no card needed. After that, the first chapter of each course stays free and full access is a one-time $39 ($19 for up to 100 founding members chosen for helpful feedback). Your progress is kept. How the beta works

Chapter 0

Readiness review

Four labeled pillars -- Linear Algebra, Probability, Statistics, Optimization -- each with an arrow rising into one shared bar labeled Machine Learning
This course asks a genuinely new question -- how do you learn a pattern from data? -- but it asks that question using math you've already built.

Four Courses, One New Question

This course builds directly on four published courses: linear-algebra, probability, statistics, and optimization. This chapter teaches, at working depth and in this course's own voice, the specific facts, identities, and conventions the chapters ahead reach for constantly — each parent course is named only as where to go for the full, rigorous treatment, never as a substitute for what's taught here. What machine learning adds is a genuinely new question none of the four asks on its own: given data, and no formula handed to you in advance, how do you find a pattern that generalizes to data you haven't seen yet?

This chapter is not a fifth course to work through, but it is a real one to read, not skim. Section 1 teaches vectors and matrices read as data: one row per example, one column per feature. Section 2 teaches Bayes' theorem in its classification role — turning a prior belief about a data point's class into a posterior belief after seeing its features. Section 3 teaches bias, variance, and the point-estimate/interval-estimate distinction, translated into a model's own predictions. Section 4 teaches gradient descent and convexity — literally what "training a model" turns out to mean, mechanically. Sections 5 and 6 teach something genuinely new: how to read NumPy array notation and pandas DataFrame notation, since every later chapter's Python examples assume both. Section 7 is this course's own map — the four kinds of problem it answers, and an honest accounting of what it teaches in full versus points you toward a later course for.

If a fact in this chapter feels completely familiar, that's the point — read it anyway for the specific way this course is going to reuse it.

  1. 0.1Vectors and matrices, read as data ~10 min
  2. 0.2Bayes' theorem and distributions, for ML ~9 min
  3. 0.3Estimation, bias, and confidence, for ML ~9 min
  4. 0.4Gradient descent and optimization, for ML ~8 min
  5. 0.5Reading NumPy array notation ~8 min
  6. 0.6Reading pandas DataFrame notation ~8 min
  7. 0.7This course's roadmap ~7 min
Start with 0.1 → Chapter quiz
Foundations of machine learning and the ML workflow →
  • No matches.