We have seen what can be learned by the perceptron algorithm ā namely, linear decision boundaries for binary classification problems.
It may also be of interest to know that the perceptron algorithm can also be used for regression with the simple modification of not applying an activation function (i.e. the sigmoid). I refer the interested reader to open another tab.
We begin with the punchline:
XOR
Now clearly, taking a ruler, your finger or positioning any straight-lined object on the above figure will not enable you to separate the blue (true) from the red (false) circles. This was also one of Marvin Minsky's arguments against further development of the Perceptron in 1963. However, with the benefit of hindsight, we shall not retire so quickly, instead we add another layer of the neurons: