External Speaker: 10am Friday Nov 9th Where: JBH359
Title: So, What Do I Do Now?
Speaker: Edna Sullivan

Abstract:
This is an extremely common question asked by many recent or upcoming
graduates, and for that matter, professionals at various stages of their
career. Nearly 30 years into a career that includes working with many
diverse organizations and applications, I propose to share my observations
and insight about characteristics of several areas of computer careers, as
well as current research into entry-level career opportunities.

Author: Edna Sullivan
Network/Database/Systems Administrator
Carlsbad Environmental Monitoring and Research Center

--------------------------------------------
Masters Project Presentation Date November 9th, 2007 Time 10:00am - 11:00 am
Place JBH 391

Title A Java Implementation of the Bayesian Data Reduction Algorithm

Candidate Dan Patterson

Advisor Dr. Turner

Committee Members Dr. Concepcion Dr. Shubert

Abstract

This project concerns the creation of the Automatic Data Reduction System,
a statistical classification application using the Bayesian Data Reduction
Algorithm (BDRA).  The algorithm was developed by Dr. Robert Lynch of the 
Naval Undersea Warfare Center and Dr. Peter Willett of the University of 
Connecticut.  The minimal set-up of the algorithm and its good prediction 
rate showed it had the potential to be developed into a commercial product.
The Office of Technology Transfer (OTTC) at California State University 
funded this project. with the computer science department to develop a 
desktop application of the BDRA.  The main goal of the project was to 
develop an accurate working implementation of the algorithm. We aimed to 
make the implementation efficient in both space and time and also allow for
it to be easily extended to be run on a distributed environment. This first
iteration only interface is the command line, but later iterations will 
develop a graphical user interface or Web front end, depending on the needs
of clients.
--------------------------------------------
Graduate Independent Study California State University San Bernardino 
Department of Computer Science
Date Friday, November 9, 2007
Time 12:00 - 12:30
Place JBH 391
Title Comparing the BDRA with PCA for Use as a Preprocessor to Artificial 
Neural Networks
Presenters Li Zhang

Abstract
Bayesian Data Reduction Algorithm (BDRA) is a statistical classification 
procedure co-invented at the University of Connecticut by Robert Lynch and
Peter Willett [1]. It performs automated probabilistic feature selection 
and classification. Artificial Neural Network (ANN) is the traditional 
approach for statistical classification.  Principle Component Analysis 
(PCA) is often used together as a preprocessing step to reduce the feature
dimensionality of the input data to improve ANN performance. In this 
presentation, I will explain the basic ideas of BDRA, PCA and ANN, and then
compare the performance of BDRA and PCA as preprocessors for neural 
networks.

Reference
[1] RS Lynch Jr, PK Willett, "Bayesian Classification and Feature Reduction
Using Uniform Dirichlet Priors", Systems, Man and Cybernetics, Part B, IEEE
Transactions on, 2003