The given ipy. provides the eigenfaces example with PCA and SVMs$.$ Assume that you have built a facial recognition model as your supervisor requested using the given PCA and

SVM settings. If you see the imported evaluation metrics in ipy, you will soon notice that the model's accuracy is in stable condition. Before running SVM$,$ the model conducted

PCA to reduce the dimensionality of features to a manageable size by setting $150$ number of components during the PCA.

In the same sampling condition $($random$\_$state of O$),$ test the evaluation metric variance of the given SVM without altering C and gamma parameters in SVMile testing

different set of n$\_$components in PCA. What happens to your facial recognition model?

In the Eigenface technique, the space of images is projected to high dimensional space; overall accuracy starts to increase as n$\_$components become lower

After testing the construction of a low$-$dimensional linear subspace that contains most of the face images possible, the recognition model with n$\_$components of $50$

performs reports overall higher accuracy than the recognition model with n$\_$components of $10$

In the Eigenface technique, the space of images is projected to low dimensional space, and overall accuracy starts to increase as n$\_$components become close to zero

After testing the construction of a low$-$dimensional non$-$linear subspace that contains most of the face images possible, the recognition model with n$\_$components of $200$

performs reports overall higher accuracy than the recognition model with n$\_$components of $100$

Even if we alter other hyperparameters of the sampling distribution, the facial recognition model would always report the same mean accuracy score over all iterations