Consider training a single perceptron $($with sign activation function$)$

to linearly separate data points into two classes C$1($shown as $1)$

and $C2($shown as $-1$