A multilayer perceptron is an instance of a feedforward neural network that only uses fully connected layers consisting of perceptrons. A perceptron is a node that takes input values and multiplies them by weights, and then passes this aggregate value to an activation function that returns a value that indicates how much this set of inputs and weights matches the pattern we are trying to find.
The multilayer perceptron can be thought of as the most basic neural network implementation. As we mentioned, all layers are fully connected, which means that there are no convolution or pooling layers. It is also a feedforward model, which means that information from backpropagation is not looped back at every step, as it is in a recurrent neural network.
Simplicity can be an asset in terms of the ease of the interpretability of the network...
