Consider what would happen if we applied a linear transformation to each data point $x^{-}|x_1,x_2|$ as follows: $x_1=a_1{x}_1+c$ and $x_2=a_2{x}_2+c,$ where $a_1,a_2,$ and $c$ are real$-$valued constants. Assume that there exists a zero crror perceptron modet for the original training set with parameters $.$

If the original data set is linearly separable with positive margin, is there always a set of parameters $w(w_0{w}_1^{\text{'}},w_2^{\text{'}})$ for a zero$-$error perceptron model on the transformed training set? Either derive the new parameter vector W$\text{'}($showing your work$),$ or explain why no such parameters exist.