Consider training a single perceptron $($with sign activation function$)$ to separate data points into two classes C$1$ and C$2.$ The samples $\{(-1,1),(1,1),(1,0)\}$ are associated with class C$1,$ and samples $\{(1,3),(3,2),(-1,-2)\}$ are associate with class C$2.$ The perceptron won't converge $($i$.$e$.$ cannot be trained$)$ when applied to these samples, why?