# CS539: Graphical Methods for Modeling, Inference, and Data Analysis

## Course Description

Graphical methods have transformed
probabilistic modeling and stand poised to revolutionize statistical
data analysis. Graphical modeling (aka Bayesian networks, belief
networks, chain graphs, undirected models, markov random fields, ...)
provide a visual representation of highly-structured probabilistic
models of compolkex domains, and have been used in medicine, science,
engineering, and the military. Efficient methods exist for both model
discovery and inference.
We will begin with a quick overview of the theory behind graphical
probabilistic modeling. We will review basic representations,
inference methods, and model discovery methods. Most of the course
will be organized around a series of case studies. Anticipated case
studies include: (1) medical diagnosis; (2) biological model
discovery; (3) web visitor behavior modeling; (4) military situation
assessment. Other topics will be explored as time permits and class
interest dictates.

Prerequisites include basic understanding of probability and
elementary statistics, together with experience and/or interest in
system modeling.

**Faculty:** Bruce D'Ambrosio, 107 Dearborn, 737 5563,
dambrosi@cs.orst.edu

**Registration Information:** 4 Units. MWF 4-5pm GBAD 103

## Class Materials

Class materials will consist of a series of
readings to be handed out in class. We will also be utilizing a number
of research and commercial software systems for graphical modeling.

### Current reading list is HERE

## Topics

- Introduction to Graphical Modeling
- Elicitation
- Prediction
- Diagnosis
- Decision-Making
- Discovery
- parameter estimation
- Structure Discovery
- Inference about causal structure

- Dynamic modeling
- Spatial modeling

*Bruce D'Ambrosio, dambrosi@cs.orst.edu *