Ponente: Prof. Ann Nicholson.
Centro de origen: Faculty of Information Technology. Monash University, Australia.

Día: Miércoles, 6 de julio.
Hora: 11h00
Lugar: salón de actos I3A

Titulo: Applications of Bayesian Networks

Resumen/Abstract:

Bayesian networks (BNs) have become a popular AI representation for reasoning under uncertainty, with successful applications in a range of domains. In this seminar I will describe two recent case studies.

The first case study looks at anomaly detection in vessel tracks using BNs.  Here we train both dynamic and static BNs with real world Automated Identication System (AIS) data, and with supplementary data. We find that the learned networks are quite easy to examine and verify despite incorporating a large number of variables, and combining the mined models improves performance in a variety of cases.

The second case study is a tool, called ABC-BN, for building, maintaining and using BNs in an environmental management application. This is an extension to a web-based application used by a state government department in Australia for monitoring over 700 threatened species and communities.  ABC-BN supports the construction of BNs built using expert elicitation around a standard template, to model the interactions between actions, threats and population and habitat factors for a species.  The tool also allows users to do what-if scenario reasoning, and is integrated with ongoing monitoring and reporting. Unlike most ecological BN modelling to date, which typically takes months or years to produce a single, often complex BN, for a specific problem, the aim was to produce a tool that would facilitate the development, maintenance and use of a large number of quite simple BNs. We describe the incremental development of ABC-BN over 3 years, with the challenges of changes to the underlying BN template, and a number of substantial additions to the tool requirements. A prototype version was evaluated in 2009 by building models for 100 species, and the completed ABC-BN has been deployed in
April 2011.