Boxi Learn › Computing › Stage 4 · Systems and Argument
One artificial neuron learns to sort
A program that learns to sort things correctly must understand them the way a person does.
A neuron multiplies each input by a weight, adds them up and says 'yes' if the total passes a threshold, so it splits the plane with a straight line; learning is nudging the weights after every mistake.
Explain to someone how this neuron gets all 16 fruits right without knowing what a fruit is. Then say why it can never get the XOR fruit all right, using the straight line.
Score 10 snacks from 0 to 10 for sweetness and for crunch, and plot them on squared paper: a circle if a family member likes it, a square if not. Can one ruler line split circles from squares? Try to find four snacks that no line can split.