Question 2. PCA on MNIST Because the number of attributes of the MNIST data set may be too big to apply kNN on it (due to the 'curse of dimensionality'), we would like to compress the images down to a smaller number of transformed attributes. Use scikit-learn to output a data set X_train_transformed and X_test_transformed, with l attributes. Here a reasonable choice of l is 10, equal to the number of labels. But you can try slightly smaller or larger values as well.