Bayesian networks
Generally, all Probabilistic Graphical Models have three basic elements that form the important sections:
- Representation: This answers the question of what does the model mean or represent. The idea is how to represent and store the probability distribution of P(X1, X2, …. Xn).
- Inference: This answers the question: given the model, how do we perform queries and get answers. This gives us the ability to infer the values of the unknown from the known evidence given the structure of the models. Motivating the main discussion points are various forms of inferences involving trade-offs between computational and correctness concerns.
- Learning: This answers the question of what model is right given the data. Learning is divided into two main parts:
- Learning the parameters given the structure and data
- Learning the structure with parameters given the data
We will use the well-known student network as an example of a Bayesian network in our discussions to illustrate the concepts...
