Skip to content
Math & AI

AI Lab

Watch a neural network learn in 3D, then deploy it to a virtual Raspberry Pi.

Start lesson 1Practice exam
Also on your phone · AI Lab
Download on theApp StoreGet it onGoogle Play
Live labLessons37Guides2
Loading 3D lab…
First lesson01 · What is a neuron?See all 37 lessons →

Report a mistake

What is wrong?
0 / 2000 · at least 10 characters
Sent along with your report
  • Page
  • About AI Lab · tinyml
Plus your browser type. Reports go to our support system; see the privacy notice.
The course

9 worlds · 37 lessons

World 1 · Signals & Neurons

FREEWhat a neuron actually is
  1. 01What is a neuron?Inputs in, one number out · 3 min
  2. 02Weights, bias, activationThe three dials of a neuron · 4 min
  3. 03From a neuron to a decisionDrawing a line · 3 min

World 2 · Learning by Falling

PROHow a network actually learns
  1. 01Loss: measuring wrongnessOne number for how bad we are · 3 minPRO
  2. 02Gradient descentRolling downhill · 4 minPRO
  3. 03The learning rateFeel it in the Loss Valley · 5 minPRO

World 3 · Deeper Minds

PROWhen one line is not enough
  1. 01When a line failsThe spiral problem · 3 minPRO
  2. 02Hidden layersComposing simple parts · 4 minPRO
  3. 03Watch it separate the spiralThe Decision Surface · 5 minPRO

World 4 · Raspberry Pi & GPIO

PROThe $4 computer you can build with
  1. 01What is a Raspberry Pi?A whole computer on one board · 3 minPRO
  2. 02GPIO & your first blinkLight an LED with Python · 4 minPRO
  3. 03Inputs & PWMButtons, and dimming with PWM · 4 minPRO
  4. 04The GPIO headerForty pins, and the two numbers each one has · 4 minPRO
  5. 05Why the LED needs a resistor3.3 volts, and what a pin can actually give · 4 minPRO
  6. 06Sequences and timingLists, loops and a traffic light · 5 minPRO
  7. 07Reading a button properlyFloating pins, pull-downs and bounce · 5 minPRO
  8. 08Making soundThe buzzer, and beeping in patterns · 3 minPRO
  9. 09Keeping stateA variable that survives the loop · 4 minPRO
  10. 10When it breaksReading errors instead of guessing · 4 minPRO

World 5 · Make It Tiny

PROShrinking a model to fit
  1. 01QuantizationFrom 32 bits to 8 · 4 minPRO
  2. 02Shrink it yourselfThe Shrinker · 5 minPRO

World 6 · Deploy to the Edge

PRORun it on a $4 chip
  1. 01RAM, flash, and latencyThe three budgets · 4 minPRO
  2. 02Flash it to the chipThe Chip · 5 minPRO

World 7 · Practice Lab

PROSix shapes a network can learn
  1. 01Two Blobs — a straight lineWhen one line is enough · 3 minPRO
  2. 02XOR — why one line failsThe problem that needs a hidden layer · 4 minPRO
  3. 03Circles — a closed boundaryWrapping one class inside another · 4 minPRO
  4. 04Two Moons — a curved splitThreading between crescents · 3 minPRO
  5. 05ReLU vs tanhThe neuron changes the path · 4 minPRO
  6. 06The optimizer mattersAdam vs Momentum on the same spiral · 4 minPRO

World 8 · A landscape you can measure

PROTwo parameters, five observations, one exact loss surface
  1. 01Build the loss landscapePRO
  2. 02Follow the gradientPRO
  3. 03When a step overshootsPRO
  4. 04A minimum is not always zeroPRO

World 9 · From six features to a point cloud

PROPCA variance and t-SNE neighbourhoods
  1. 01Keep directions that carry variancePRO
  2. 02A map can distort neighboursPRO
  3. 03Optimize a 3D t-SNE mapPRO
  4. 04Change the neighbourhood scalePRO
Read first

Guides for AI Lab

All guides →
Explainer
Logic GatesLogic Gates Explained: AND, OR, NOT, XOR and NANDHow logic gates turn 1s and 0s into decisions, how to read truth tables, why NAND can build every other gate, and how De Morgan's laws connect them.4 min read
Explainer
AI LabWhat Is an Artificial Neuron? Weights, Bias, ActivationA single artificial neuron takes numbers in and gives one number out. See how weights, bias and activation work and how a neuron draws a decision line.4 min read