Machine Learning in Production: Developing and Optimizing Data Science Workflows and Applications

12. Dimensional Reduction and Latent Variable Models

12.1 Introduction

Now that you have the tools for exploring graphical models, we’ll cover a few useful ones. We’ll start with factor analysis, which finds application in the social science and in recommender systems. We’ll move on to a related model, principal components analysis, and explain how it’s useful for solving the collinearity problem in multiple regression. We’ll end with ICA, which is good for separating signals that have blended together. We’ll explain its application on some psychometric data.

One thing all of these models have in common is that they’re latent variable models. That means in addition to the measured variables, there are some unobserved variables that underlie ...