Introduction
In the last chapter we saw that a single neuron can only draw straight lines, and that adding a hidden layer solves problems like XOR. A multilayer perceptron (MLP) is exactly that: layers of neurons stacked one after another. It's the classic "plain" neural network, and every other architecture — CNNs, RNNs, Transformers — contains MLP-style layers inside it.
Think of a factory assembly line. Raw materials (the inputs) enter at one end. Each station (layer) transforms what it receives a little and passes it on. At the end, a finished product (the prediction) comes out. Each station's machinery has settings (weights) that training will tune.