Department of Management and Marketing
Center for the Management of Science and Technology (CMOST)
|
|
Workshop
Computing Simple Explanations from Numerical Data
by University of Texas at Dallas, Bioxsys GmbH, Berlin, Germany & SullTrue Inc., Richardson, Texas
May 3 and 4, 2007
Administrative Science Building, Room 106 |
|
| Instructor | |
| Overview | |
| Registration | |
| Schedule | |
| Contact | |
| Campus Map | |
| Acknowledgment | |
| FLYER |
sponsored by
UAH Rotorcraft Systems Engineering and Simulation Center
UAH Center for the Management of Science and Technology
organized by
Department of Management and Marketing
College of Administrative Science
University of Alabama in Huntsville
|
Dr. Klaus Truemper, Professor Emeritus of Computer Science at the University of Texas at Dallas, is an expert in the design of intelligent systems to solve complicated problems. An internationally known scholar, he is the author of three academic books and a number of academic papers in Computer Science, Operations Research, and Mathematics.
He has developed several intelligent software systems out of his theoretical work for solving complicated practical problems such as handwriting interpretation, text processing, traffic control, medical diagnosis, and data mining. Notably, Dr. Truemper and his research partners have reported new findings in cancer research as a result of applying his software systems to cancer patient data.
Dr. Truemper has frequently been invited to give workshops, tutorials and seminars on the design of logic-based intelligent system in universities and research centers over several countries in recent years.
His research work has been supported by grants from government and private agencies in the U.S., Italy, and Germany, including the National Science Foundation and Department of Defense. In addition to his academic work, he is also very experienced in consulting on solving practical problems for businesses such as Texas Instruments and Southwest Airlines.
Dr. Truemper was invited to give a seminar at UAH, September 2006, as part of the Distinguished Speaker Series at UAH.
The two day workshop consists of six sessions designed to introduce participants to a new technique for constructing simple explanations from numerical data. This workshop will particularly focus on the very difficult situation where comparatively few data records with numerous data elements per record are given. Applications of this new technique arise in a variety of areas such as engineering, testing, bioinformatics, economics, finance, and medicine. The techniques used in these sessions are applicable to analysis involving rotorcraft systems and the requirement to construct safety and reliability predictions from numerical data. It is applicable to condition-based maintenance and reliability-centered maintenance.
Key software for the technique is the Leibniz System. One module of the software handles the derivation of explanations from data and thus handles the construction of empirical models. Altogether, the software supports the construction, evaluation, and operation of intelligent systems based on empirical as well as normative models.
Participants can choose to attend either the first day or both days of the workshop. Day one will focus on obtaining explanations from numerical data and the Leibniz System Software modules. Day two will focus on the mathematical theory underlying the Leibniz system and will cover both basic and advanced techniques of data mining and logic reasoning.
The workshop will make use of the following materials:
Registration is closed. All the seats have been filled up. You will be placed in the waiting list if registering now.
Day 1: Thursday May 3, 2007
The first day consists of three sessions. The morning sessions includes an application-oriented discussion on Obtaining Explanations from Numerical Data. The second session includes an explanation of the Modules of the Leibniz System Software. The final session is demonstration on the use of the software.
Check-in and Refreshments
9:30 - 10:00am
Session 1
10:00 - 11:00am Obtaining Explanations from Numerical Data
11:00 - 11:30am Question and Answer
Summary: The explanation problem demands the construction of two explanations that tell why the numerical records of one population differ from those of a second population. The explanations must be simple enough that humans can comprehend them for decision making.
A new multi-step method called EXARP solves the explanation problem. The steps carry out discretization of data, selection of important factors, construction of explanations, and validation. While these terms are familiar in machine learning, the actual processes are quite different from prior ones. In particular, in each step a so-called alternative random process (ARP) is introduced that attempts to distort information or disrupt the computing process. Appropriate action by the solution algorithms prevents the ARPs from reaching these goals.
The lecture covers various examples in bioinformatics, economics, engineering, finance, and medicine. They show EXARP to be very effective even when data sets are very small.
Session 2
1:30 - 2:30pm: Explanation Modules of the Leibniz System
2:30 - 3:00pm Question and Answer
Summary: Discretization and learning logic modules of the Leibniz System, input and output files, computation of important factors and explanations, treatment of continuous effects.
The session discusses use of the modules, including preparation of input, selection of parameter files, and evaluation of output files including statistical validation tests.
Session 3
3:00 - 4:00pm: Lab Session Demonstrating the Computation of Explanations.
Summary: The demonstration includes installation of the Leibniz System, preparation of input files, selection of processing steps, and evaluation of output files.
Day 2: Friday May 4, 2007
The second day also consists of three sessions. The morning session covers the Basic Techniques for Computation of Explanations. The second session, section 5, focuses on Advanced Techniques for Computation of Explanations. The final session is an additional Lab Session on the use of the software.
Session 4
10:00 - 11:00am Algorithms for Computation of Explanations
11:00 - 11:30am Question and Answer
Summary: Alternate random processes, discretization, learning of logic formulas, important factors, construction of explanations, and validation.
This is a blackboard session where the mathematical theory underlying the Leibniz System is presented in a classroom atmosphere. The goal is not a polished but hard-to-understand presentation, rather an interactive presentation where the audience is encouraged to ask questions for clarification of statements and results.
Session 5
1:30 - 2:30pm: Advanced Explanation Topics
2:30 - 3:00am Question and Answer
Summary: Outliers, effect intervals, importance functions, costs of tests and optimal explanations, and normative constraints imposed on explanations.
This session is a continuation of the morning program. Here, advanced concepts and related algorithms underlying the Leibniz System are discussed.
Session 6
3:00 - 4:00pm: Lab Session
Summary: This is a second lab session demonstrating the particular aspects of the computation of explanations.
This workshop is made possible by the financial support of the UAH Rotorcraft Systems Engineering and Simulation Center, Center for the Management of Science and Technology, and Department of Management and Marketing. Dr. Klaus Truemper is especially acknowledged for his kindly agreeing to offer the workshop at UAH. |
Simulation Center