Multilayer Perceptrons: Modern Feed Forward Networks

A multilayer perceptron is a special case of a feedforward neural network where every layer is a fully connected layer, and in some definitions the number of nodes in each layer is the same. Further, in many definitions the activation function across hidden layers is the same. One should generally use the multilayer perceptron when one knows very little about the structure of the problem. One uses neural networks to do feature learning. Using fully connected layers only, which defines an MLP, is a way of learning structure rather than imposing it. While one can increase the depth and width of the network, that simply increases the flexibility in function approximation. Features are entirely learned.