Artificial intelligenceLesson 1 of 1 in the course Reasoning under uncertainty

Bayesian Networks: Inference, Structure, and Learning

5 flashcards · 5 quiz questions · 5 match pairs · 5 fill-in sentences · English

Save course
Lesson progress
0 / 5

Your progress is kept in this browser only.Sign up to save it

Summary

A Bayesian network is a way to represent uncertain relationships among variables using a graph and probability distributions. This lesson introduces the graph structure and shows how local relationships can represent a full joint probability distribution.

A Bayesian network is a probabilistic graphical model. Its nodes represent random variables, such as observable measurements, hidden states, unknown parameters, or hypotheses. A directed edge from one node to another represents a direct conditional dependency in the model.

All flashcards

What is a Markov blanket in a Bayesian network?
A node’s parents, children, and the other parents of its children; given these, the node is independent of the rest of the network.
What is a collider in a directed graph?
A pattern in which two arrows point into the same node, such as A→M←BA\to M\leftarrow B.
What does variable elimination do?
It sums or integrates out non-query variables while combining probability factors.
What is expectation-maximization used for in parameter learning?
It alternates between estimating hidden-variable values and improving parameter estimates.
What is shrinkage in a hierarchical model?
Individual estimates are pulled toward a shared group-level value by information pooled across the group.

Quiz questions

  1. What does the local Markov property state for a node in a Bayesian network?Answer: It is independent of its non-descendants given its parents.
  2. In probabilistic inference, what is done with variables that are not part of the query?Answer: Their possible values are summed out or, for continuous variables, integrated out.
  3. In a causal network, what happens to a node’s mechanism under an intervention that sets its value?Answer: Its incoming arrows are cut and its usual probability mechanism is replaced by the set value.
  4. Why does treewidth help predict the cost of exact inference?Answer: The worst-case cost grows exponentially with treewidth.
  5. What is the main goal of structure learning for a Bayesian network?Answer: To discover the graph structure, and possibly its parameters, from data.

Based on the Wikipedia article Bayesian network, licensed CC BY-SA 4.0. Adapted into a summary, flashcards and quizzes.

More in Artificial intelligence

All Artificial intelligence lessons
Artificial intelligenceA* Search: Pathfinding, Heuristics, Guarantees, and Variants5 flashcards · 5 quiz questions · 4 gamesPlay free →Artificial intelligencePDDL Core Structure, Version History, and Planning Extensions5 flashcards · 5 quiz questions · 4 gamesPlay free →Artificial intelligenceDescription Logic: Knowledge Bases, ALC Semantics, and Reasoning5 flashcards · 5 quiz questions · 4 gamesPlay free →