Chapter 6. Probabilistic Graph Modeling

Probabilistic graph models (PGMs), also known as graph models, capture the relationship between different variables and represent the probability distributions. PGMs capture joint probability distributions and can be used to answer different queries and make inferences that allow us to make predictions on unseen data. PGMs have the great advantage of capturing domain knowledge of experts and the causal relationship between variables to model systems. PGMs represent the structure and they can capture knowledge in a representational framework that makes it easier to share and understand the domain and models. PGMs capture the uncertainty or the probabilistic nature very well and are thus very useful in applications that need scoring or uncertainty-based approaches. PGMs are used in a wide variety of applications that use machine learning such as applications to domains of language processing, text mining and information extraction, computer...