Design of Logic-based Intelligent Systems

Wednesday, September 29, to Friday, October 1, 2004

Prof. Dr. Klaus Truemper
Department of Computer Science
University of Texas at Dallas

Location: Konrad-Zuse-Zentrum Berlin
Takustr. 7, 14195 Berlin-Dahlem
lecture hall (room 2005)

At this workshop, jointly organized by the European Graduate Program "Combinatorics, Geometry, Computation" and the DFG Research Center "Mathematics for key technologies", Prof. Dr.Klaus Truemper (University of Texas at Dallas) will introduce his new book "Design of Logic-based Intelligent Systems" and report about some highlights of this research area. Location of the workshop, time schedule and program are listed below. All lectures will be held in English.

Time schedule:

There will be two talks each day followed by a software tutorial.

Times: 13:00-14:00 and 14:30-15:30 (talks), 15:30-open end (software tutorial)
Venue: Konrad-Zuse-Zentrum, Takustr. 7, 14195 Berlin-Dahlem, lecture hall (room 2005)

A summary of the whole lecture series and brief abstracts of the presentations are given below.

Summary:

The workshop covers the design and implementation of logic-based intelligent systems. Such a system accomplishes feats that, when achieved by humans, require significant intelligence. Examples are medical diagnosis, processing of natural language, and supervision of complex processes. The material is based on the book "Design of Logic-based Intelligent Systems," published by John Wiley & Sons in May 2004. Most algorithms of the book have been implemented in software called the Leibniz System. Each lecture links the discussion of theory with the relevant portion of the Leibniz System. Participants of the workshop have the opportunity to explore use of the software after each lecture.

Lecture 1: Introduction to Intelligent Systems
(Wednesday, Sept. 29, 13:00 - 14:00)

Definition of intelligent systems, levels of thinking, expert systems and intelligent agents as special cases. Logic techniques such as production rules, Prolog, propositional logic, first-order logic, fuzzy logic, and Bayesian networks. Extension of propositional logic. Deduction by theorem proving via problem SAT. Optimization via cost minimization problem MINSAT. Other kinds of reasoning: abduction, induction, nonmonotonic reasoning, default reasoning, metareasoning
[Slides (pdf)]

Lecture 2: Basic Formulation and Reasoning Techniques
(Wednesday, Sept. 29, 14:30 - 15:30)

Decision pyramid of variables and clauses. Finding redundant and inconsistent clauses via problem MINCLS UNSAT. Validation of formulations, explanation of decisions, accelerated theorem proving for related questions. Uncertainty of clauses, proof at a likelihood level, existence of a solution at a confidence level
[Slides (pdf)]

Lecture 3: Learning Logic Formulas from Data
(Thursday, Sept. 30, 13:00 - 14:00)

Separation of two data sets by min/max or optimized logic formulas, handling of additional logic constraints. Collections of formulas and voting, probability distributions of vote totals, classification control. Transformation of rational data and set data to logic data, handling of unknown information by values Absent and Unavailable.
[Slides (pdf)]

Lecture 4: Nonmonotonicity, Incompleteness, Futility
(Thursday, Sept. 30, 14:30 - 15:30)

Nonmonotonicity and incompleteness of formulations, including extension to uncertainty. Futility of theorem proving and problems Q-ALL SAT, Q-MINFIX UNSAT, and Q-MINFIX SAT.
[Slides (pdf)]

Lecture 5: Question-and-answer Processes
(Friday, Oct. 1, 13:00 - 14:00)

Question-and-answer processes: selection of goals and tests, computation of optimal plans. Fully automated construction of expert systems using learned logic formulas, which is an example of thinking at the second level.
[Slides (pdf)]

Lecture 6: Examples of Intelligent Systems
(Friday, Oct. 1, 14:30 - 15:30)

Correctness of machine design, music composition assistant, management of hazardous materials, traffic control, credit rating, deciding word sense, differential medical diagnosis.
[Slides (pdf)]