Skip to main content

Normally one node/layer applies linear fitting of the the input to the hypothesis, in other words uses linear function ($y = ax + b$). Adding layers chains liner functions, potentially allowing fitting higher order functions. A great explanation can be found here.