One neuron, one straight boundary

Every position on the map stands for a possible pair of inputs: one colour value and one size value.

Feed any pair into the neuron. If the weighted total stays below the threshold, the neuron answers dangerous. If the total reaches the threshold, it answers safe.

The answer changes at an edge. Draw that edge on the map and it becomes a line. This line is called the .

Every input pair on one side gets one answer. Every pair on the other side gets the other.

This neuron only multiplies, adds, and checks one threshold, so its decision boundary is always straight.

Changing the weights tilts the line. Changing the threshold moves it. Training searches for settings that place the line as well as possible.