Vol 24. Year 1999 Abstracts

No. 3

V. Mousseau, J. Meric:

Empirical violation of the third aletrnative independance principle: implications for multicriteria decision aid

Abstract

Multiple criteria decision aiding methodologies rely on basic principles concerning the consistency of decision maker's (DM's) preferences. Among these principles, the Third Alternativ Invariance (TAI) principle states that elicited preferences on two alternatives should not be afected by the presence or availability of a third alternative. Empirical statements on cognitive limitations of decision makers tends to contradict this principle. Our work aims at analysing the behavior of DMs when answering binary and ternary choice questions. Results show a violation of the TAI principle. Such results ma be explained by a shift in the reference points considered by decision makers in the binary and ternary choice questions. Implications for preference elicitation techniques in multiple criteria decision aid are discussed.

J. Józefowska, G. Waligóra, J. Weglarz:

Parallel tabu search algorithms for scheduling independent jobs on parallel machines to minimize the makespan

Abstract

In this paper implementations of parallel Tabu Search algorithms for solving the problem of scheduling nonpreemtable, independent jobs on parallel, identical machines to minimize the makespan are presented. Four different types of parallel Tabu Search based on Michael J. Flynn computer architectural classification scheme are considered. A computational experiment is described performed in two different parallel computer architectures: multiprocessor scalar system with shared memory and scalable supercomputing system with distributed memory. A comparison of results obtained for sequential versions of the algorithm as well as parallel ones in both the architectures is presented. Some conclusions and final remarks are included.

A. N. Giovanis, C.H. Skiadas:

The method of local linearization for the solution of stochastic innovation diffusion models

Abstract

This paper examines the performance of a numerical method, known as local linearization method, for the solution of stochastic innovation diffusion models. The method is applied to stochastic Bass innovation diffusion model and a methodology for the comparison of the discretized stochastic model with the respective deterministic local trend model are provided. Finally the local linearized stochastic model is applied to real data where the superiority of the proposed methodology is illustrated.

G. Sarker:

A new concept learning system for image recognition with generalisation control

Abstract

A new Concept Learning System (CLS) for regular image recognition has been designed and developed in the present paper. The images are preprocessed and each compressed image constitute an internal version of the image instance. The instances of different image classes provided to the system, one class at a time, allows the system to extract the essential features which constitute the concept for each class. These features along with the class names provided by the supervisor during learning is housed in the knowledge base of the system. The system has the unique feature to recognize unknown objects with varying degree of generalization as per requirement by the end user.

No. 4

S.G. Tzafestas, S.N. Raptis, N.A. Maschos:

A hybrid pricing expert system based on fuzzy reasoning

Abstract

Regardless of the increasingly more sophisticated and precise mathematical models formulated in the fields of economics, a persistent gap remains between the predictions and the economic reality. The uncertainty and complexity of the involved phenomena fails to be encapsulated into . precise. models. With the most important part of the related information stemming from human experts, it is more than reasonable to employ models that are capable of hosting and reasoning with such knowledge. It is argued that the fuzzy logic and fuzzy reasoning framework is perfectly suited for such an approach since it provides the required expressiveness for the knowledge representation and the respective inference mechanisms to efficiently deal with the pricing problem. This argument is discussed and verified with the use of the FUZZYPRICE system that is designed and implemented to assist pricing decision-making. The results obtained are then compared to the well-known PRICE-STRAT system to demonstrate the effectiveness of the current approach.

J. Blazewicz, G. Pawlak, M. Sterna, B. Walter, M. Zawirski:

Heuristic algorithm for schedule optimization in FMS environment

Abstract

The paper concerns an existing flexible manufacturing system containing two dedicated machines with three magazines of raw material and the transportation system in the form of the robot with a linear routing. Firstly, a mathematical model of the system has been developed. In general, there are batches of identical parts to be produced in the system for which the technological routes are defined and which constitute a job-shop environment. Additionally, the set-up times for batches and particular operations are taken into account. Fm has been developed. In general, there are batches of identical parts to be produced in the system for which the technological routes are defined and which constitute a job-shop environment. Additionally, the set-up times for batches and particular operations are taken into account. For this model a heuristic  scheduling algorithm has been constructed. Computational experiments and simulation studies validated the model and the algorithm proposed.

R. Spin-Andrade, E. Fernandez: "A fuzzy logic model for bargaining"

Abstract
In this work a fuzzy logic model for bargaining is presented. This model "translates" a group of assumptions obtained from a frequently consulted non mathematical literature, which reflects the experience of prestigious authors about the topic of negotiation, establishing a simple and consensual rationality in contrast with previous approaches. Once this fuzzy model is built, it can be used in connection with the theory of Von Neumann and Morgestern for n-personal games, becoming in a new solution paradigm for these. The model implemented on a MATLAB program can be used as a tool to ease agreement among parts and also to predict the result of the bargaining process.

D. Schweigert: "Ordered graphs and minimal spanning trees"

Abstract
The set of edges of a graph G=(V,E) are considered as an ordered set (E;<). Concerning this order we generalize the graph-theoretical optimizations problems like the problem to find a minimal spanning tree. Using linear extensions L(E) of the ordered set (E;<) one can apply the well known algorithms. We show that for every minimal spanning tree T there exists a linear extension L such that T can be found by a well-known algorithm like Prim's or Kruskal's.