Beginner's Guide to Deep Learning

We concluded that a single perceptron with step function could implement each one of the fundamental logical functions: NOT, AND, and OR.
They are called fundamental because of any logical function, irrespective of how complex it is can be obtained by a combination of these three. We can infer that, if we appropriately connect the three perceptrons we just built, we can implement any logical function!

# XOR Multi-Layer Perceptron
XOR is not linearly separable, so we need a mechanism to build a network such that it creates a boundary to separate the two classes. Let’s see how can we build a network of fundamental logical perceptrons so that it implements the XOR function.

From Perceptron to Neural Network