Introduction to Machine Learning — the course

A free bilingual machine-learning course you can follow on your own: short videos, open readings, a classifier you train yourself, and a progress tracker. No programming.

Lessons, videos and readings

• ML1 — Rules, examples, patterns — Nobody can write a rule for every face, every handwriting or every customer. Machine learning replaces the rule with enough examples. This first lesson gives you that idea and the honest limits that come with it.
  Elements of AI — Chapter 1: What is AI? — University of Helsinki · Machine Learning Crash Course — Framing — Google
• ML1 — The five words you need — Data, feature, label, model, prediction. Five words that let you follow any meeting, any vendor pitch and any later lesson without nodding at things you do not understand.
  Machine Learning Glossary — Google · Elements of AI — Chapter 2: AI problem solving — University of Helsinki
• ML2 — Prediction, grouping, or neither — Almost every honest machine-learning project is one of two shapes. The third category — decisions about a person’s job, credit, health or discipline — is where we stop and keep a human responsible.
  Supervised and unsupervised learning, explained simply — Google · Elements of AI — Chapter 4: Machine learning — University of Helsinki
• ML3 — From your notebook to a table — A model can only learn from what you wrote down. This lesson turns real sales, stock, attendance or request records into rows and columns, and shows what dirty data does to a result.
  Data preparation and feature quality — Google · Build a clean data table in a spreadsheet — LES Institute
• ML4 — Train your first classifier — Thirty photos per class, a browser, no code. You train a working image classifier, then you deliberately break it so you can see how a model fails.
  Teachable Machine — train an image model — Google · Training and test sets — why they must differ — Google
• ML5 — Why 95% accuracy can be worthless — Accuracy is the number vendors quote and the number that hides the most. You learn to ask what the errors are, who they fall on, and what each one costs.
  Classification: accuracy, precision, recall — Google · Fairness and bias in datasets — Google
• ML6 — Model, rule, or nothing — and the one-page brief — The final lesson turns the course into a decision. Most of the time the answer is a rule and a spreadsheet, and saying so is the professional result. Then you write the brief that a developer or a vendor can work from.
  Deciding whether machine learning is the right tool — Google · Elements of AI — Chapter 6: Implications — University of Helsinki