In a series of observations, two features f_1 and f_2 are measured. When four measurements of (f_1, f_2) are (-1, 2), (2, -1), (1, 1), and (-2, -2), what is the covariance matrix when performing PC A of the observations? Why does PCA compute eigenvectors and eigenvalues of the covariance matrix of the observations, and not those of a different matrix