T1
Aspectos Teóricos y Prácticos de Lógica Difusa
Claudio M. Held, Ph.D.
|
Abstract: This tutorial is targeted to an audience with no prior experience in Fuzzy Logic, as an introduction to the subject. It is divided in two parts. The first part will take place on Monday, December 15th, 14:00-15:30 hrs., in a room to be informed soon enough (we hope), and the second part, immediately after the break, 16:00 to 17:00 hrs, in the Computer Intelligence Lab. The talk will be given in Spanish, unless the audience wishes otherwise. The presentation (slides) is mostly in English. The first part of the tutorial has a lecture format. A main motivation to study Fuzzy Sets is that there are several situations in the real world where object classes have membership categories which are not precisely defined. One convenient way to formalize these are Fuzzy Sets. A short introduction includes the discussion of some theoretical aspects, such as membership grades, possibilistic distributions and operations on Fuzzy Sets and Fuzzy Logic. The two main practical applications of Fuzzy Logic are Fuzzy Logic Control (FLC) and Decision-Making Support (DMS). In FLC the emphasis is given to the Mamdani approach, i.e. modeling the operator. However, our main goal in the examples part is to clarify DMS applications. A major part of the lecture is devoted to show different examples, aimed at providing a view on the wide scope of the different possible applications. The examples which are more extensively described refer to a basketball trainer decision support system, handwritten digits recognition, classification of sleep stages in infants and fraud detection in telecommunications. In some cases the Fuzzy Logic application lays at the very center of the solution, in others it provides a convenient addition to contribute to a solution. The second part of the tutorial aims at providing a hands-on experience. Two simulation experiments will be carried out by the participants. The first one is a noise-cancelling filter which uses a FLC. It is implemented in Simulink (Matlab). Several parameters of the filters can be manipulated, and their effects visualized in the simulation software. The second experiment mimics an expert system, and is implemented in Matlab. The goal is to identify a certain animal (within a limited universe) by a series of questions to the user, whose answers guide the search for the solution. The participants will be asked to look into the rule base, make some modifications, and observe the consequences. In this type of applications the man-machine interface is very important for a successful acceptance by the end user. To provide a satisfying hands-on experience in the simulation lab, it is the second part of the tutorial which limits the maximum number of participants allowed. |