Probabilistic graphical models with OpenMarkov, an open-source tool
Presenter:
Francisco Javier Díez, PhDDept. Artificial Intelligence, UNED, Madrid, Spain
Description
This course will introduce probabilistic graphical models (PGMs), such as Bayesian networks, influence diagrams and decision analysis networks, and discuss their advantages over traditional techniques; for example, influence diagrams and decision analysis networks are equivalent to decision trees containing thousands of branches, Markov influence diagrams can model state-transition problems without multiplying the number of states and decision analysis networks can evaluate large models with unordered decisions. OpenMarkov, an open-source tool, allows building PGMs for complex problems using a graphical user interface, without writing any code, such as spreadsheet formulas, macros or functions. For beginners, the course will be an introduction to medical diagnosis and decision making. For participants who have already conducted cost-effectiveness analyses it will be interesting to see that building and evaluating a PGM is easier, faster and less error-prone than building and debugging an equivalent model using a spreadsheet, a (Markov) decision tree or a programming language, such as R, MATLAB or C++. Participants are invited to bring their own laptops with OpenMarkov installed.
Content
- Outline and schedule (PDF)
- Slides (PDF, 13 MB) [updated: Nov. 2 2019, 23:55]
- Hands-on exercises (PDF)
Attendants must install version 0.3.2 of OpenMarkov.
We recommend printing the outline and the exercices (on paper) and practicing with OpenMarkov's tutorial before attending the course.
Additional information
Short bio
F. J. Díez is full professor of artificial intelligence at UNED, the largest Spanish university. In his PhD thesis he built DIAVAL, one of the first Bayesian networks for medicine. He has been principal investigator in several national and international projects and published his work in some the most relevant journals of AI, including Artificial Intelligence and Artificial Intelligence in Medicine, and health decision analysis, such as Medical Decision Making and Pharmacoeconomics. He is the leader of a research group that built OpenMarkov, an open-source tool for PGMs, especially tailored to medicine, which has been used in more than 30 countries. He has been teaching probabilistic graphical models to computer science students for 25 years. He is the director of a modular program that has taught around 2,000 health professionals since 1996.
Selected publications (in the recommended order of reading):
- M. Arias, F. J. Díez (2014) The problem of embedded decision nodes in cost-effectiveness decision trees. Pharmacoeconomics 32:1141-1145.
- M. Arias, F. J. Díez (2015) Cost-effectiveness analysis with influence diagrams. Methods of Information in Medicine 54:353-358.
- F. J. Díez, M. Yebra, I. Bermejo, et al. (2017) Markov influence diagrams: a graphical tool for cost-effectiveness analysis. Medical Decision Making 37:183-195.
- F. J. Díez, M. Luque, I. Bermejo (2018) Decision analysis networks. International Journal of Approximate Reasoning 96:1-17.
- J. Pérez-Martín, I. Bermejo, F. J. Díez (2019) Evaluation of Markov models with discontinuities. Medical Decision Making 39:414-420.
- F. J. Díez, M. Luque, M. Arias, J. Pérez-Martín (2019) Cost-effectiveness analysis with unordered decisions.
- J. Pérez-Martín, M. A. Artaso, F. J. Díez (2017) Cost-effectiveness of pediatric bilateral cochlear implantation in Spain. Laryngoscope 127:2866–2872.
- M. Luque, F. J. Díez, C. Disdier (2016) Optimal sequence of tests for the mediastinal staging of non-small cell lung cancer. BMC Medical Informatics and Decision Making 16:1-14.