 |
Session 1: Diagnostic Systems
(August 7, 2003, 8:30 – 9:45 AM)
These papers extend the standard approach to Bayesian Network diagnostic modeling by addressing large-scale models, combining expert and statistical sources of knowledge, information-based selection of tests and probes, and applicability to beyond fault identification, to process control and predictive maintenance.
- B. Kappen, W. Wiegerinck, E. Akay, J. Neijt, A. van Beek:
 Promedas: A clinical diagnostic decision support system
(176kb)
- O. Kipersztok, G. Provan:
 A Framework for Diagnostic Inference of Commercial Aircraft Systems
(33kb)
- I. Rish, M. Brodie, N. Odintsova, S. Ma, G. Grabarnik:
Problem Diagnosis in Distributed Systems using Active Probing
(577kb)
- G. Weidl, A. L. Madsen, E. Dahlquist:
Object Oriented Bayesian Networks for Industrial Process Operation
(482)
Session 2: Analysis and Testing of Models
(August 7, 2003, 10:00 – 11:00 AM)
Characteristic of all these papers is a second order concern, to judge the quality of a model’s conclusions.
Session 3: Game Theory Applications
(August 7, 2003, 11:15 AM – 12:00 PM)
These papers consider problems with multiple decision makers with competitive or collaborative values, such as techniques for generation of non-manipulable auction mechanisms.
Session 4: Multimedia Data
(August 7, 2003, 1:00 – 1:45 PM)
Issues with non-traditional data, in both textural and geographic domains.
Session 5: Active Sensing
(August 7, 2003, 2:00 – 3:00 PM)
These models make decisions at each stage that affect their structure, e.g. by changing which sensors to use, or by use of switching nodes. These models make decisions at each stage that affect their structure, e.g. by changing which sensors to use, or by use of switching nodes. All but Mahoney represent DBN models.
Session 6: Dynamic Models
(August 7, 2003, 3:15 – 4:00 PM)
These papers model dynamic situations, by a dynamic or temporal technique such as modifying the model over time. They are distinguished from the DBN and MDP techniques that use stages to model time explicitly.
Session 7: Planning and Scheduling
(August 7, 2003, 4:15 – 5:30 PM)
These models use related probabilistic approaches, including Markov Decision Processes (MDP) and Dynamic Bayesian Networks (DBNs) to solve decision problems with multiple decisions or a sequence of decisions.
|
