An activation function is used to introduce non-linearity to a network. This allows us to model a class label / score that varies non-linearly with independent variables. Non-linear means the output cannot be replicated from a linear combination of inputs, this allows the model to learn complex mapp ...READ MORE...