This book focuses on the theoretical as well as practical uses of probabilistic graphical models, commonly known as PGM. This is a technique in machine learning in which we use the probability distribution over different variables to learn the model. In this book, we have discussed the different types of networks that can be constructed and the various algorithms for doing inference or predictions over these models. We have added examples wherever possible to make the concepts easier to understand. We also have code examples to promote understanding the concepts more effectively and working on real-life problems.
Chapter 1, Bayesian Network Fundamentals, discusses Bayesian networks (a type of graphical model), its representation, and the independence conditions that this type of network implies.
Chapter 2, Markov Network Fundamentals, discusses the other type of graphical model known as Markov network, its representation, and the independence conditions implied by it.
Chapter 3, Inference – Asking Questions to Models, discusses the various exact inference techniques used in graphical models to predict over newer data points.
Chapter 4, Approximate Inference, discusses the various methods for doing approximate inference in graphical models. As doing exact inference in the case of many real-life problems is computationally very expensive, approximate methods give us a faster way to do inference in such problems.
Chapter 5, Model Learning – Parameter Estimation in Bayesian Networks, discusses the various methods to learn a Bayesian network using data points that we have observed. This chapter also discusses the various methods of learning the network structure with observed data.
Chapter 6, Model Learning – Parameter Estimation in Markov Networks, discusses various methods for learning parameters and network structure in the case of Markov networks.
Chapter 7, Specialized Models, discusses some special cases in Bayesian and Markov models that are very widely used in real-life problems, such as Naive Bayes, Hidden Markov models, and others.