You are developing an ordinary least squaresbased predictor and have been given a data set with n examples with d real-valued features. You decide to expand the feature space by adding as well as multiplying pairs of features; that is, you create d * (d - 1) additional features. Your method performs a straightforward formulation to find the optimal set of weights. Is there something you should be worried about? Yes. my addition-based features are guaranteed to lead to an unstable solution. Yes, my multiplication-based features are guaranteed to lead to an unstable solution. No. I proceed uninterrupted to compute the solution vector. No. l proceed to apply the standard solution. In fact. I am very happy that my extra features are giving me a chance to find a non-linear decision boundary