A hidden layer transforms input values inside a neural network. Its values are hidden because they are neither supplied as inputs nor observed as outputs; the output layer turns the final transformation into a prediction. section 11.2.1
Each node combines weighted inputs, adds a bias, and applies an activation function. Together, the node outputs form a learned representation: a new numerical description of the input, not a copy of it. section 11.2.1
Image representations
In an image network, one layer can represent simple visual patterns, while later layers combine them into shapes and objects. section 11.2.1










