Four points. One useful lesson.
Meet XOR. Follow the actual weights, activations, and outputs of a tiny network, with the chain rule close at hand.
Explore XOR Recorded introduction · training truth tableA small network. A deeper understanding.
Follow a prediction from input to output. Replay a learning experiment. Then change a weight and see what the mathematics predicts.
p = sigmoid(tanh(xW₁ + b₁)W₂ + b₂)01 / EXPERIMENTS
Start with a small truth table, then explore patterns that need a curved boundary.
Meet XOR. Follow the actual weights, activations, and outputs of a tiny network, with the chain rule close at hand.
Explore XOR Recorded introduction · training truth tableReplay the moons experiment. Inspect predictions, compare a linear baseline, and take your first temporary gradient step.
Explore moons Interactive · replay, decisions & gradientsExplore concentric classes and see what hidden neurons add. Check whether a better fit carries over to held-out examples.
Explore circles Interactive · replay, decisions & gradientsTHE GRADIENT LAB
Choose any weight or bias. Compare backpropagation with a numerical derivative, inspect the loss curve, and try a step. Overshot? Undo it and reduce the learning rate.
Open the gradient labθ′ = θ − η∇J
Up to 50 temporary steps, with undo and reset. Saved models and evaluation results stay fixed.
02 / FOLLOW THE MATHEMATICS
Read the explanation, make a prediction, and check your reasoning against the worked answers.
Weights, matrix shapes, forward propagation, and the chain rule.
Data splits, standardization, mini-batches, Adam, L2, and early stopping.
Thresholds, error costs, ranking, calibration, and input sensitivity.
Gradient checks, loss slices, tangents, and a complete update by hand.