In the case of SVM for example, it cannot separate data if it is not linearly separable. By using the square of Feature1 as the third parameter, separation becomes possible as shown below.
