A joint distribution over many variables can always be factored via the product rule, but that factorisation is only interesting when some variables genuinely don't depend on others. Directed graphical models (Bayesian networks) and undirected graphical models (Markov random fields) are two visual languages for making exactly those dependency assumptions explicit — a picture that simultaneously specifies a factorisation of the joint distribution and lets you read off conditional independence statements without any algebra.
In this lesson
- Learning objectives
- Directed graphical models (Bayesian networks)
- Reading conditional independence off a graph
- Undirected graphical models (Markov random fields)
- Worked example
Before you start
- Overview: Probabilistic and Graphical Models
- Generative Classifiers: Gaussian Discriminant Analysis and Naive Bayes
The full lesson, with worked examples and exercise solutions, is for members.
Read the full lessonNew to SelfMadeQuant.com? See membership options or browse the Financial Machine Learning syllabus.