Let D be an n×d data matrix, and y be an n-dimensional column vector containing the dependent variables of linear regression. The regularized solution to linear regression predicts the dependent variables of a test instance Z using the following equation:

Prediction(Z) = Z W = Z(DTD + λI)

−1DT y Here, the vectors Z and W are treated as 1×d and d×1 matrices, respectively. Show using the result of Exercise 1, how you can write the above prediction purely in terms of similarities between training points or between Z and training points.