Chapter 1 · The dawn of AI (1943–1990)1969
The XOR problem
A single perceptron cannot learn XOR because no straight line separates it; this stalled neural networks.
Chapter 1 · The dawn of AI (1943–1990)1969
A single perceptron cannot learn XOR because no straight line separates it; this stalled neural networks.
One vertical line: two dots on the wrong side.
XOR means “one or the other, but not both”. The dots at opposite corners share an answer, so every straight line leaves at least one dot on the wrong side.
XOR gives 1 when exactly one of its two inputs is 1. So 0,0 gives 0; 1,0 gives 1; 0,1 gives 1; and 1,1 gives 0.
Coral dots should be 1 and cyan dots 0. Move the line, then add a second layer. Dashed rings are dots on the wrong side. A teaching version: you move the lines by hand.
Three out of four is the best one line can ever do. Try as hard as you like, then switch to two lines.
What is XOR?
Why can a single perceptron not learn XOR?
What solves it?