Artificial Intelligence By Example

XOR and linearly separable models

In the academic world, like the private world, competition exists. Such a situation took place in 1969. Minsky and Papert published Perceptrons. They proved mathematically that a perceptron could not solve an XOR function. Fortunately, today the perceptron and its neocognitron version form the core model for neural networking.  

One might be tempted to think, So what? However, the entire field of neural networks relies on solving problems such as this to classify patterns. Without pattern classification, images, sounds, and words mean nothing to a machine.