BMAW 2015
Annual Bayesian Modeling

Applications Workshop

16 July 2015 - Workshop Day


Amsterdam, The Netherlands

Workshop Agenda

Technical paper authors were assigned 20 minutes to present their work plus 10 minutes for questions. Position paper authors will present their work as posters during the breaks (coffee break and lunch break).

Bayesian network structure learning (BNSL) is the problem of finding a Bayesian network structure which best explains a given dataset. Score-based learning, a widely-used technique for solving BNSL, assigns a score to each network structure. The score measures the goodness of fit of that network to the data; commonly-used scores are based on, for example, Bayesian posterior likelihoods or minimum description length principles. Solving BNSL optimally is known to be NP-hard. Nevertheless, in the last decade, a variety of algorithms have been proposed which guarantee to find a network structure that optimizes a given scoring function. The goal of this work is to better understand the empirical behavior of these algorithms.

In order to finish within a reasonable amount of time, most of the exact algorithms (safely) prune the search space using a variety of heuristics, such as branch-and-bound search. After briefly introducing several of the exact algorithms, I will discuss problem-dependent characteristics which affect the efficacy of the different heuristics. Empirical results show that, despite the complexity of the algorithms, machine learning techniques based on these characteristics can often be used to accurately predict the algorithms' running times.

Within BNSL, the score typically reflects fit to a training dataset; however, it is well-known that a model may fit a training dataset very well but generalize poorly to unseen data. Thus, it is not clear that finding a network which optimizes a scoring function is worthwhile. The second part of this talk will focus on a study which compares exact and local search techniques. In some scenarios, algorithms such as greedy hill climbing or the polynomial-time Chow-Liu algorithm suffice; for more complex datasets, though, the exact algorithms consistently produce better networks.

Very recently, several algorithms have been developed for learning bounded-treewidth networks, in which exact inference is guaranteed to be efficient. This talk will conclude with discussion on preliminary results which suggest that the generalization performance of bounded-treewidth is similar to that of more complex structures.

Note: A tutorial (on the first day of the main UAI conference, July 12) by Changhe Yuan and James Cussens will focus on BNSL algorithms. Consequently, this talk will only include a brief overview of the algorithms.

This talk largely pulls from the following references.

Malone, B.; Kangas, K.; Järvisalo, M.; Koivisto, M. & Myllymäki, P. "Predicting the Hardness of Learning Bayesian Networks." Proceedings of the 28th AAAI Conference on Artificial Intelligence, 2014.
Malone, B.; Järvisalo, M. & Myllymäki, P. "Impact of Learning Strategies on the Quality of Bayesian Networks: An Empirical Evaluation." Proceedings of the 31st Conference on Uncertainty in Artificial Intelligence, 2015.
Berg, J.; Järvisalo, M. & Malone, B. "Learning Optimal Bounded Treewidth Bayesian Networks via Maximum Satisfiability." Proceedings of the 17th International Conference on Artificial Intelligence and Statistics, 2014.

09:15 - 10:30

Opening Session

09:15 - 09:30


09:30 - 10:30

Invited Talk: Empirical Investigations into Bayesian Network Structure Learning Algorithms

Dr. Brandon Malone
paper, presentation

10:30 - 11:00



11:00 - 12:30

Risk, Ivestment, and Tool

11:00 - 11:30

Project Cost, Benefit and Risk Analysis using Bayesian Networks

Barbaros Yet, Anthony Constantinou, Norman Fenton, Martin Neil, Eike Luedeling and Keith Shepherd
paper, presentation

11:30 - 12:00

Bayesian Optimisation of Gated Bayesian Networks for Algorithmic Trading

Marcus Bendtsen
paper, presentation

12:00 - 12:30

A Tool for Visualising the output of a DBN for fog forecasting

Tal Boneh, Xuhui Zhang, Ann Nicholson and Kevin Korb
paper, presentation

12:30 - 14:30



14:30 - 16:00


14:30 - 15:00

An IRT-based Parameterization for Conditional Probability Tables

Russell Almond
paper, presentation

15:00 - 15:30

Bayesian network models for adaptive testing

Martin Plajner and Jirka Vomlel
paper, presentation

15:30 - 16:00

Computer Adaptive Testing Using the Same-Decision Probability

Suming Chen, Arthur Choi and Adnan Darwiche
paper, presentation

16:00 - 16:30



16:30 - 17:30

Vehicle and Aircraft Applications

16:30 - 17:00

Influence diagrams for the optimization of a vehicle speed profile

Václav Kratochvíl and Jirka Vomlel
paper, presentation

17:00 - 17:30

Bayesian Predictive Modelling: Application to Aircraft Short-Term Conflict Alert System

Vitaly Schetinin, Livija Jakaite and Wojtek Krzanowski
paper, presentation

17:30 - 18:00

Planning for BMAW 2016

17:30 - 18:00

Final discussion