Full definition

Mathematical function applied to the output of an artificial neuron that determines if it should be activated, introducing non-linearity to the network.

Example in a business context

ReLU, Sigmoid and Tanh are common activation functions that allow the network to learn complex patterns.