University of Dayton
Computer Science Department
Computer
Science Colloquium
Academic Year 2002 - 2003
All colloquia events will
meet in the following locations and times:
Talk: 4:30 –
5:30pm in 215Miriam Hall
Coffee and Cookies: 5:30-6:00pm in the Computer
Science Conference Room
Dates: the following Tuesdays-- 10/1/2002, 10/15/2002, 11/19/2002, 4/4/2003, 4/11/2003
Tuesday, October 1,
2002 ***
Miriam
Hall 215; 4:30 pm
***Note:
New Date
Evolutionary
Optimization of Voting Classifiers
Dr.
Dale E. Courte
University
of Dayton
Cincinnati, Ohio
ABSTRACT
When searching for solutions to classification problems involving high dimensionality of input and/or multiple target classes, data preprocessing techniques can be useful in determining an appropriate feature set through feature extraction. However, determination of appropriate features is very dependent on the classification scheme being used. The determination of appropriate preprocessing is not always straightforward. In addition, the choice of an appropriate classification scheme itself is not always clear. This dissertation describes the Cooperative Hybrid Architecture for Multi-class Problems, or CHAMP. Using optimization techniques provided by various approaches to evolutionary computation, the CHAMP model provides a framework for hybrid evolutionary systems that optimize a cooperating set of pairs, each consisting of a feature extractor and a classifier. Supervised learning allows appropriate feature extractors and classifiers to be selected from pre-defined families of each. This selection is accomplished through configurations encoded in the genetic material of the evolutionary system. The classifiers participate in a voting scheme that allows them to cooperatively yield the final classification of an input pattern. Feedback in the form of fitness measurements determines refinements to the configurations, affecting changes in the feature extraction, feature selection, and classification procedures during the evolutionary process. The viability of the architecture is demonstrated incrementally through a series of experiments, each chosen to highlight specific capabilities, using both synthetic and real-world problems.
Dr. Dale Courte
Dale Courte spent 24 years on the campus of Wright State University in a variety of roles. Positions held include Computer Science Instructor, Systems Adminisrator, and Associate Director of Computing and Telecommunication Services. Most recently he was a Graduate Reseach Assistant and DAGSI Scholar completing his Ph.D. in Computer Science and Engineering. In the fall of 2002, he joined the Computer Science Faculty at the University of Dayton. His research applies the techniques of evolutionary computation to the field of pattern recognition.
Tuesday, October 15, 2002
Miriam Hall 215; 4:30 pm
Abduction as a Reasoning Process
Dr. Richard Fox
Department of Mathematics and Computer Science
Northern Kentucky University,
Cincinnati,
Ohio
ABSTRACT
Artificial Intelligence investigates mechanisms whereby automated reasoning systems can be constructed. All too often, these reasoning system use a reasoning process that is implied by the given programming language (or programming shell) used, such as backward chaining as found in Prolog, forward chaining as found in OPS5, or resolution theorem proving as found in logic-oriented tools. A better way to construct a reasoning system is to base the
problem solving strategy on the reasoning processes found when investigating how humans solve the problem. One common form of reasoning found in a large variety of problems is abduction, or abductive inference. This is a reasoning process where one attempts to explain the cause of some behavior or situation through hypothesization, evaluation and composition. Many AI researchers have investigated abduction over the past 20 years and have found it to yield a much more focused form of reasoning than other approaches. This talk will introduce abduction as a process and briefly examine several different implementations of abductive inference. The talk will then focus on one particular strategy, developed at The Ohio State University's Laboratory for Artificial Intelligence Research (LAIR) and describe several very disparate problem solving systems that have applied this strategy. In will conclude with a brief look at how this strategy might be applied to other compositional problems such as design and decision making.
Dr. Richard Fox
Dr. Richard Fox is an Associate Professor in the Department of Mathematics and Computer Science at Northern Kentucky University. This is his second year at this position. From 1992 to 2001, Dr. Fox taught in the Department of Computer Science at the University of Texas - Pan American. He received his Ph.D. in Computer and Information Sciences from The Ohio State University in 1992, specializing in Artificial Intelligence.
Friday, November 22, 2002
Miriam Hall 213; 3:00pm
Java: An
Eventful Approach
Dr. Douglas Troy,
Professor and Chair
Computer
Science & Systems Analysis Dept.
Miami
University, Oxford, OH
ABSTRACT
This semester at Miami University we are experimenting with a new approach to teaching introductory Java programming. The approach was developed by faculty at the Williams College Computer Science Department for teaching S1. It is an objects-first approach, uses graphics and animation heavily, introduces event-driven programming from the start, and even introduces concurrency. A Java class library supplied by the Williams College faculty simplifies catching of mouse events and drawing and manipulating simple graphics objects. For example, students do not have to know about Java interfaces, the Java event model, or the paint method to use event-driven programming and to draw graphics.
For lab and programming assignments, students write simple video game-like programs that are fun, motivating, and challenging. The first half of the course relies on the mouse as the only source of input. In the latter part of the course, we progress to using some of the basic Java GUI (AWT and Swing) components where we introduce the standard Java event model.
In this talk, I will present the approach and many of the laboratory exercises and programming assignments that we have used this semester.
Friday, April 4, 2003
Miriam Hall 215; 3:00 pm
NASA's
Next Generation of Autonomous Explorers
NASA
Ames Research Center
Moffett
Field, California
ABSTRACT
NASA's
future missions to acquire scientific knowledge about deep space require
increased capabilities for autonomous decision making for surface explorers and
spacecraft. To build these capabilities, NASA is investing in research and
technology in Artificial Intelligence and related fields in Computer Science.
This talk will provide an overview of the technology challenges in areas such
as automated planning, control, fault management, and coordination that must be
overcome in order to achieve the level of robust autonomy required to meet
NASA's deep space mission requirements.
Robert
Morris is a researcher in planning and scheduling in the Computational Sciences
Division at NASA Ames Research Center, NASA's Center of Excellence in
Information Technology. He is also program manager in NASA's Intelligent
Systems Program in the area of Automated Reasoning. Before joining NASA
permanently in 1999, Dr. Morris was a professor of Computer Science at the
Florida Institute Of Technology. He received a Ph.D. in Philosophy at Indiana
University and an M.S. in Computer Science at Wright State University. The main
focus of his research is on constraint-based approaches to planning and
scheduling.
Friday, April 11, 2003
Miriam Hall 215; 3:00 pm
Bioinformatics
Biology
Department
University
of Dayton
Dayton, OH
45469
ABSTRACT
To be provided at a later date.
Past colloquia:
·
Fall 2001 and Spring 2002
Back to
Computer Science Homepage CPS Home