Chapter 1 · The dawn of AI (1943–1990)1958
The perceptron
Rosenblatt's perceptron learns its own weights by nudging them after each mistake.
Chapter 1 · The dawn of AI (1943–1990)1958
Rosenblatt's perceptron learns its own weights by nudging them after each mistake.
4 dots on the wrong side. Nudge the line.
Dashed rings mark dots on the wrong side. Each mistake nudges the line a little, and the line slowly finds the gap between the colours.
A perceptron is a neuron with weights and a bias. Together they draw a straight line: dots on one side get one answer, dots on the other side get the other.
Tap the square to add a dot of the chosen colour; tap a dot to remove it. Then press Step to fix one mistake, or Auto to keep going.
Some dots are on the wrong side. Press Step or Auto.
This is the classic perceptron rule on two inputs, a simplified teaching version.
When does a perceptron change its weights?
What shape does a single perceptron draw to split two groups?
What happens if the colours criss-cross so no line splits them?