Three ways to learn
Learning with answers, learning without answers, and learning by trial and error.
Not every problem comes with an answer key. Depending on what data you have, machine learning comes in three main flavours.
🏷️ Supervised learning
Every example comes with the right answer, called a label. Photo → "cat". House details → price.
Predicting a category is classification. Predicting a number is regression.
Most practical ML today is supervised.
🔍 Unsupervised learning
No labels at all. The model looks for structure by itself.
Example: grouping customers into types by shopping habits (clustering), without anyone saying what the types are.
🎮 Reinforcement learning
An agent takes actions in an environment and gets rewards or penalties.
No one tells it the right move; it learns which actions pay off over time. This is how AlphaGo learned to beat world champions at Go.
You have 50,000 photos of skin spots, each labelled by a doctor as harmless or worrying. Which kind of learning fits?
🪄 Self-supervised: the secret behind chatbots
A clever trick: make labels out of the data itself. Hide the next word of a sentence and ask the model to guess it.
The internet becomes one giant answer key. This is how large language models are pre-trained.
A robot learns to walk by trying moves and getting points for staying upright. That's…
✨ Before you drift off
- Supervised: learn from labelled examples (classification or regression).
- Unsupervised: find hidden structure without labels.
- Reinforcement: learn actions from rewards.
- Self-supervised: make labels from the data itself, which is how LLMs pre-train.
📚 Go deeper (free & open)
- Elements of AI, chapter 4: Machine learning ↗ · University of Helsinki & MinnaLearn
- Kaggle Learn: Intro to Machine Learning ↗ · Kaggle · Short free hands-on micro-courses