Partial Least Squares and Principal Component Analysis

Latent variable methods are very powerful tools that can be utilized to extract features from multi-variable datasets. Some of these methods are Principal Component Analysis and Partial least squares. These algorithms define a subspace of your dataset which has a lower dimensionality, but still captures most of the information contained in the dataset. In other words, they find linear combinations of the columns in a dataset that have the maximum variability content. These new variable (Columns) are called the latent variables. following PDF describes these methods in more details and provides visual examples of their application