國家理論科學研究中心學術演講
 NCTS Seminar on Probability

with Biological Applications

Speaker:      Dr. Han-Shen Huang  黃漢申

                    (Institute of Information Science, Academia Sinica)

Time: :          (I) 13:30-15:00, April 7, 2005

                    (II) 13:00~15:00, April 14, 2005 -- Postponed to Apr. 21, 2005

Topic:           (I) Introduction to Constructing Bayesian Networks

                    (II) Introduction to Using Bayesian Networks

Abstract:

(I) We will introduce the background knowledge for applying Bayesian networks to real-world domains. A Bayesian network is a graphical model that encodes the joint distribution of a set of random variables, and we can query the model for marginal distribution of a subset of random variables given some evidences. It can be used to model the real-world domains with uncertainty based on well-known and well-understood probability theory. One of the advantages of Bayesian networks is that the model itself is a white box to those who are familiar with probability. Therefore, unlike the black-box models, a Bayesian network can be constructed by the cooperation of domain experts and machine learning techniques. We will focus on the cores to construct Bayesian network models.

(II) We will demonstrate how to take advantage of a Bayesian network software to quickly construct and make inference on your Bayesian network models. The software we use is Hugin (http://www.hugin.com), one of the most famous Bayesian network softwares. We hope that every attendant should bring a laptop with Hugin Lite installed on it (http://www.hugin.com/Products_Services/Products/Demo/). Knowledge of constructing Bayesian networks is required.

Room 213,

Science One Building,

National Chiao Tung University