1. Learning Bayesian Networks for Complex Relational Data
Tutorial Website
Learning Bayesian Networks for Complex Relational Data
Tutorial Introduction
Presenters: Oliver Schulte, Ted Kirkpatrick
School of Computing Science
Simon Fraser University
Vancouver-Burnaby, Canada
Markov Logic Networks can be addressed too
total time: 220 min, 30 min break

2. Overview
Learning Bayesian Networks for Complex Relational Data

3. The Short Story
Many organizations keep their data in a relational database.
We describe methods for learning a Bayesian network for data in a relational database.
Simultaneous joint statistical analysis of multiple interrelated tables.
Questions:
What does the output mean?
How can you use it?
What are the statistical issues in learning the Bayesian networks?
What are the computational challenges?
Learning Bayesian Networks for Complex Relational Data

4. Questions Considered
Semantics: how do you interpret a relational/first-order Bayesian network?
How can you use it?
What are the statistical challenges for learning?
What are the computational challenges for learning?
statistical issues in learning multi-relational Bayesian networks
Learning Bayesian Networks for Complex Relational Data

5. Motivation
Learning Bayesian Networks for Complex Relational Data

6. Database Management Systems
Maintain data in linked tables.
Structured Query Language (SQL) allows fast data retrieval.
E.g., find all movie ratings > 4 where the user is a woman.
Multi-billion dollar industry, $Bn15+ in 2006.
IBM, Microsoft, Oracle, SAP, Peoplesoft.
Learning Bayesian Networks for Complex Relational Data

7. Beyond storing and retrieving data
Much new interest in analyzing databases.
Data Mining.
Data Warehousing.
Business Intelligence.
Predictive Analytics.
Learning Bayesian Networks for Complex Relational Data

8. Unifying Logic and Probability
Fundamental Question in AI: how to combine logic and probability and learning?
Statistical-Relational Learning
Domingos (U of W, CS): “Logic handles complexity, probability represents uncertainty.”
Recent survey paper by Stuart Russell
Learning Bayesian Networks for Complex Relational Data

9. Query Examples
Learning Bayesian Networks for Complex Relational Data

10. Sample Queries
Inference in a Bayesian network computes answers to probabilistic queries
A Bayesian network for relational data can answer relational probabilistic queries
We give some examples of relational and nonrelational queries
Learning Bayesian Networks for Complex Relational Data

11. Single-Table Queries (Not relational)
Query
English Paraphrase
P(Drama(Movie) = T|RunTime(Movie) = Long)
The probability that a movie is a drama, given that it is long.
P(Country(Actor) = U.S.|gender(Actor)=W)
The probability that an actor is from the US, given that her gender is woman.

12. Cross-Table Queries (Movies)
Query
English Paraphrase
Positive relationship
P(Drama(Movie) = T|
RunTime(Movie) = Long, ActsIn(Movie,”brad pitt”),
ActsIn(Movie,”julie delp”),
Country(“julie delp” = France))
The probability that a movie is from the US, given that it is long, and given that Brad Pitt and Julie Delp have appeared in it and Julie Delp is from France.
Negative relationship
RunTime(Movie) = Long,
ActsIn(Movie,”brad pitt”),
not ActsIn(Movie,”juliette binoche”)
Country(“juliette binoche” = France))
The probability that the movie named Movie is from the US, given that it is long, and given that Brad Pitt has appeared in it, and Juliette Binoche has not appeared in it and is from France.
For the last query, the instance space can also be just since without a relationship specified, a random movie is independent of a random actor.

13. Cross-Table Queries (Actors)
Query
English Paraphrase
Positive relationship
P(Country(Actor) = U.S.|
gender(Actor)=W, ActsIn(“hate”,Actor), RunTime(“hate”)=short)
The probability that an actor is from the US, given that she is a woman, and given that she appeared in the movie “hate”.
Negative relationship
P(Country(Actor) = U.S.|gender(Actor)=W,
not ActsIn(“hate”,Actor),
RunTime(“hate”)=short)
The probability that the actor named Actor is from the US, given that she is a woman, and given that she appeared in the long movie Movie, and did not appear in the short movie “hate”.

14. Motivating Applications
The ability to answer relational probabilistic queries has supported a number of successful applications. For example:
Relational Query Optimization
Information Extraction (DeepDive)
Ontology Matching
Entity Resolution
Link-based classification
Anomaly detection/exception mining

15. Tutorial Approach Our tutorial is a survey of issues, not of systems
We give references to different systems
Discuss a range of issues but only a single model class (Bayesian networks)
Most concepts generalize to log-linear models for relational data.
Focus on the new challenges of learning Bayesian networks with relational data, compared to traditional iid data
Illustrate challenges and solutions with a running example
in particular, generalizes to Markov Logic networks
Kimmig, A.; Mihalkova, L. & Getoor, L. (2014), 'Lifted graphical models: a survey', Machine Learning,
Sutton, C. & McCallum, A. (2007), An Introduction to Conditional Random Fields For Relational Learning’
Introduction to Statistical Relational Learning', MIT Press, , pp

16. Tutorial Plan Relational Data First-Order Bayesian networks
Parameter Learning for First-Order BNs
30 min break
Structure Learning for First-Order BNs
Link-Based Classification using a First-Order BN
Relational Anomaly Detection using a First-Order BN
Learning Bayesian Networks for Complex Relational Data