Unsupervised learning algorithms learn the properties of data on their own without explicit human intervention or labeling. Typically within the AI field, unsupervised learning technique learn the probability distribution that generated a dataset. These algorithms, such as autoencoders (we will visit these later in the book), are useful for a variety of tasks where we simply don't know important information about our data that would allow us to use traditional supervised techniques.
PCA is an unsupervised method for feature extraction. It combines input variables in such a way that we can drop those variables that provide the least amount of information to us. Afterwards, we are left with new variables that ...
