16.4 Advantages of Structured Modeling

• Allow us to dramatically reduce the cost of representing probability distributions as well as learning and inference.

• Sampling accelerated in the case of directed model.

• Allow us to explicitly seperate representation of knowledge from learning of knowledge or inference given existing knowledge. Makes our model easier to model or debug.

  • We can design, analyze and evaluating learning algorithms and inference algorithmsthat are applicable to broad class of grapgs.
  • Independently, wen can design models that capture the relationships we believe are important in our data.
  • Then combine different algorithms and structures.