CMU-CS-05-147
Computer Science Department
School of Computer Science, Carnegie Mellon University



CMU-CS-05-147

Exploiting Parameter Domain Knowledge for Learning
in Bayesian Networks

Radu Stefan Niculescu

July 2005

Ph.D. Thesis

CMU-CS-05-147.pdf


Keywords: Constraint optimization, domain knowledge, graphical models


The task of learning models for many real-world problems requires researchers to incorporate problem Domain Knowledge into the learning algorithms because there is rarely enough training data to enable accurate learning of the structures and underlying relationships in the problem. Domain Knowledge comes in many forms. Domain Knowledge about relevance of variables (Feature Selection) can help us ignore certain variables when building our model. Domain Knowledge specifying conditional independencies among variables can guide our search over possible model structures. This thesis presents a theoretical framework for incorporating a different kind of knowledge into learning algorithms for Bayesian Networks: Domain Knowledge about relationships among parameters.

We develop a unified framework for incorporating general Parameter Domain Knowledge constraints in learning procedures for Bayesian Networks by formulating this as a constrained optimization problem. We solve this problem using iterative algorithms based on Newton-Raphson method for approximating the solutions of a system of equations. We approach learning from both a frequentist and a Bayesian point of view, from both complete and incomplete data.

We also derive closed form solutions for our estimators for several types of Parameter Domain Knowledge: parameter sharing, as well as sharing properties of groups of parameters (sum sharing and ratio sharing). While models like Module Networks, Dynamic Bayes Nets and Context Specific Independence models share parameters at either conditional probability table or conditional distribution (within one table) level, our framework is more flexible, allowing sharing at parameter level, across conditional distributions of different lengths and across different conditional probability tables. Other results include several formal guarantees about our estimators and methods for automatically learning domain knowledge.

To validate our theory, we carry out experiments showing the benefits of taking advantage of domain knowledge for modelling the fMRI signal during a cognitive task. Additional experiments on synthetic data are also performed.

132 pages


Return to: SCS Technical Report Collection
School of Computer Science

This page maintained by reports@cs.cmu.edu

COMPUTER SCIENCE TECHNICAL REPORT ABSTRACTS
CMU-CS-05-146
Computer Science Department
School of Computer Science, Carnegie Mellon University



CMU-CS-05-146

Emodis—An End-based Network Monitoring
and Diagnosis System

Ningning Hu, Peter Steenkiste

June 2005

CMU-CS-05-146.ps
CMU-CS-05-146.pdf


Keywords: Architecture, monitoring, diagnosis, measurement


Network monitoring and diagnosis capabilities are critical for the seamless operation of a network. ISPs use sophisticated systems to routinely monitor and diagnose their networks, but end users do not have such capabilities. To address this problem, we develop Emodis—a network monitoring and diagnosis system. In this paper, we describe the architecture and the software components of Emodis. Like other end-user oriented network monitoring systems, Emodis is deployed on a diverse set of Internet nodes, so it shares common requirements such as security and robustness with these systems. However, the focus of Emodis is on route-sensitive path metrics such as available bandwidth and packet loss rate, resulting in two unique characteristics: (1) it implements a variety of measurement techniques, including sophisticated bandwidth measurement techniques, but hides many technical details from end users; (2) it implements a scheduling algorithm to synchronize the measurements from different vantage points, which relieves end users from complicated network measurement management.

15 pages


Return to: SCS Technical Report Collection
School of Computer Science

This page maintained by reports@cs.cmu.edu