The activation function is the function of an artificial neuron in a neural networkthat provides outputs based on inputs. This function graphically illustrates their range. It is located "at the end" of the neural structure and can be likened to the axon of a biological neuron.
There are linear activation functions (maintaining a constant) and nonlinear activation functions (creating more variation), from which the neural network is subsequently constructed.
Typical examples of activation functions include:
Identity function (do nothing, output is a linear combination),
Step function (send a pulse (ON) if the value of the linear combination is greater than 0, otherwise do nothing (OFF))
Sigmoid function (a milder version of the step function).