01A single decision
Start with the mechanism
Perceptron Lab
Watch one learning unit inspect handwriting, add evidence and adjust after a mistake.
Open the labFour interactive labs · One place to explore
See how a machine learns.
Change the examples, adjust the model, and watch what happens. Explore four ways machines learn, from a single decision to an agent finding its way through a rink.
✳Try.
Observe.
Understand.
Observe.
Understand.
Examples
Patterns
Decisions
Practice
01 · Pick a questionChoose the idea you want to understand.
02 · Change one thingUse the essential controls, then explore more settings.
03 · Look for a surpriseCompare what you expected with what the model does.
Choose your starting point
The labs
Each lab works on its own. Start anywhere, or follow the suggested path from examples to practice.
02Learning from data
Compare two approaches
Weather Lab
Fit a curve to studied days, then group weather patterns without giving the machine answers.
Open the lab03Layers and learning
Build on a single unit
Student Lab
Train a neural network on an invented class and see what its layers learn to recognise.
Open the lab04Learning by doing
Change the kind of feedback
Rink Lab
Help a skater learn from attempts, rewards and mistakes as it tries to score.
Open the labThe bigger picture
Different ways to learn. Shared questions.
What information does the learner get? What changes with practice? How can you tell if it learned something useful? Use the concept map to carry those questions between labs.
Explore the concept mapFor teaching
Bring the labs into class.
Find a suggested sequence, discussion prompts and a quick guide to what each lab shows.
Open educator guideFor offline use
Keep a copy.
Each lab is available as a self-contained HTML download that runs from a laptop without internet.
Get the offline labs