Bayesian Networks
Transcription
Bayesian Networks
THE RENEWABLE ENERGY RESOURCES – EUREGIO PROJECT Bayesian Networks Probabilistic Graphical Modeling • Applied mathematics, • uncertainity and complexity Vilem Berka, vilem.berka@fsv.cvut.cz Czech Technical University Prague, Montan Universitaet Leoben, 2006 Vilem Berka, CTU Prague – Montan Universitaet Leoben Page 1 © Vilem Berka THE RENEWABLE ENERGY RESOURCES – EUREGIO PROJECT Bayesian Networks – Definition (Wikipedia) A Bayesian network is a representation of the joint distribution over all the variables represented by nodes in the graph. Let the variables be X(1), ..., X(n). Let parents(A) be the parents of the node A. Then the joint distribution for X(1) through X(n) is represented as the product of the probability distributions for i = 1 to n. If X has no parents, its probability distribution is said to be unconditional, otherwise it is conditional. Vilem Berka, CTU Prague – Montan Universitaet Leoben Page 1 © Vilem Berka THE RENEWABLE ENERGY RESOURCES – EUREGIO PROJECT Bayesian Networks (st-andrews.ac.uk) Node – Edges graph types anne.smith@st-andrews.ac.uk Vilem Berka, CTU Prague – Montan Universitaet Leoben Page 1 © Vilem Berka THE RENEWABLE ENERGY RESOURCES – EUREGIO PROJECT Bayesian Networks (st-andrews.ac.uk) Graphical representation of a joint probability distribution, representing dependence and conditional independence relationships. Vilem Berka, CTU Prague – Montan Universitaet Leoben Page 1 © Vilem Berka THE RENEWABLE ENERGY RESOURCES – EUREGIO PROJECT Bayesian Networks (st-andrews.ac.uk) Limitation of the BN - equivalence class (Functional networks) Dynamic Bayesian networks (DBNs) – Functional loop over time. Vilem Berka, CTU Prague – Montan Universitaet Leoben Page 1 © Vilem Berka THE RENEWABLE ENERGY RESOURCES – EUREGIO PROJECT Bayesian Networks (st-andrews.ac.uk) Vilem Berka, CTU Prague – Montan Universitaet Leoben Page 1 © Vilem Berka THE RENEWABLE ENERGY RESOURCES – EUREGIO PROJECT Bayesian Networks - Example (Kevin Murphy) University of British Columbia murphyk@cs.ubc.ca Vilem Berka, CTU Prague – Montan Universitaet Leoben Page 1 © Vilem Berka THE RENEWABLE ENERGY RESOURCES – EUREGIO PROJECT Bayesian Networks – Applications Medical diagnostics and probability estimations Bayesian LAB – knowledge modeling, data mining Market simulator Network diagnostic and Repair Automatic speech recognition engines Decision making AI gaming SW Microsoft: Intelligent user interfaces. Office Assistant since version 97. Vilem Berka, CTU Prague – Montan Universitaet Leoben Page 1 © Vilem Berka THE RENEWABLE ENERGY RESOURCES – EUREGIO PROJECT Computer models in use c rre o C m he t of s n tio el od Simulation model Simulation run Interpretation Real world Abstraction, modeling Evaluation Results Experiment Vilem Berka, CTU Prague – Montan Universitaet Leoben Page 1 © Vilem Berka THE RENEWABLE ENERGY RESOURCES – EUREGIO PROJECT Bayesian Networks (BayesiaLAB) Download the Tutorial at: http://www.bayesia.com/GB/produits/bLab/BLabTour/BLabTour.php Vilem Berka, CTU Prague – Montan Universitaet Leoben Page 1 © Vilem Berka THE RENEWABLE ENERGY RESOURCES – EUREGIO PROJECT Bayesian Networks 5 reasons to use • Brainstorming, communication • Probabilistic relations with significant impact • Understanding the data • Ability to test What-if scenarios • Similarity rules applications Vilem Berka, CTU Prague – Montan Universitaet Leoben Page 1 © Vilem Berka THE RENEWABLE ENERGY RESOURCES – EUREGIO PROJECT Bayesian Network editors • Bayesia LAB trialware • Microsoft Research Download at: http://research.microsoft.com/adapt/MSBNx Vilem Berka, CTU Prague – Montan Universitaet Leoben Page 1 © Vilem Berka THE RENEWABLE ENERGY RESOURCES – EUREGIO PROJECT Microsoft research – Bayesian evaluation Demonstration – automotive example Vilem Berka, CTU Prague – Montan Universitaet Leoben Page 1 © Vilem Berka