Earlier, we realized that the dimensionality of an image is equal to the number of pixels in it. So, there is no way to visualize our 43-dimensional MNIST dataset. It is true that we can display each digit separately, yet we cannot see where each image falls in our feature space. This is important to understand the classifier's decision boundaries. Furthermore, an estimator's memory requirements grow in proportion to the number of features in the training data. As a result, we need a way to reduce the number of features in our data to deal with the aforementioned issues.
In this section, we are going to discover two dimensionality-reduction algorithms: Principal Component Analysis (PCA) and Neighborhood Component Analysis (NCA). After explaining them, we will use them to visualize the MNIST dataset and generate additional samples to add to our training set. Finally, we will also use feature selection algorithms to remove...
