Catálogo de publicaciones - libros

In this paper, we propose a new approach to deal with beliefs by supposing that a rational agent has a degree of trustiness for each information it disposes of. We propose (i) a new framework for dealing with imprecise beliefs and desires; (ii) two algorithms for updating the mental state of an agent in this new setting; (iii) three ways for comparing the resulting fuzzy set of desires and (iv) two postulates which the goal election function must obey.

The standard approach to information entropy applied to partitions of a universe is equivalently formulated as the entropy of the corresponding crisp identity resolutions, interpreted as crisp granulations, by the corresponding characteristic functionals. Moreover, in this crisp context the co–entropy notion is introduced. The extension to the case of fuzzy identity resolutions, a particular case of fuzzy granulation, is studied.

This paper presents a possibilistic linear programming model for solving the blending and multi-mode, multi-period distribution planning problem with imprecise transportation, blending and storage costs. The solution procedure uses the strategy of simultaneously minimizing the most possible value of the imprecise total costs, maximizing the possibility of obtaining lower total costs, minimizing the risk of obtaining higher total costs. An illustration with a data set from a realistic situation is included to demonstrate the effectiveness of the proposed model.

Multi-adjoint logic programming represents an extremely flexible attempt for introducing fuzzy logic into logic programming (LP). In this setting, the execution of a goal w.r.t. a given program is done in two separate phases. During the operational one, admissible steps are systematically applied in a similar way to classical resolution steps in pure LP, thus returning an expression where all atoms have been exploited. This last expression is then interpreted under a given lattice during the so called interpretive phase. In declarative programming, it is usual to estimate the computational effort needed to execute a goal by simply counting the number of steps required to reach their solutions. In this paper, we show that although this method seems to be acceptable during the operational phase, it becomes inappropriate when considering the interpretive one. Moreover, we propose a more refined (interpretive) cost measure which fairly models in a much more realistic way the computational (special interpretive) a given goal.

This paper presents a computability theorem for fixed points of multi-valued functions defined on multilattices, which is later used in order to obtain conditions which ensure that the immediate consequence operator computes minimal models of multilattice-based logic programs in at most ω iterations.

Possibilistic abductive reasoning is particularly suited for diagnostic problem solving affected by uncertainty. Being a Knowledge-Based approach, it requires a Knowledge Base consisting in a map of causal dependencies between failures (or anomalies) and their effects (symptoms). Possibilistic Causal Networks are an effective formalism for knowledge representation within this applicative field, but are affected by different issues. This paper is focused on the importance of a proper management of explicit contextual information and of the addition of a temporal framework to traditional Possibilistic Causal Networks for the improvement of diagnostic process performances. The necessary modifications to the knowledge representation formalism and to the learning approach are presented together with a brief description of an applicative test case for the concepts here discussed.

Fuzzy Inductive Reasoning (FIR) methodology is a very powerful tool for creating a mixed qualitative-quantitative model of any dynamical system by using its input and output signals. One of the key issue of this methodology is the creation of the mask , i.e. a matrix that contains the causal dependencies among the signals of the systems for particular time steps. This paper describes the ARMS – Automatic Reconstruction of the Mask Scheme – methodology that gives the opportunity of creating a sub-optimal mask with very good performances without an exhaustive search in the space of all the possibilities. This methodology has been validated on a wide class of dynamical system (from LTI systems to chaotic time series) and it has been compared to other methods proposed in literature.

Universal approximation is the basis of theoretical research and practical application of fuzzy systems. The ability of fuzzy models to model static information has been successfully proven and tested, while on the other hand their limitations in simultaneously modelling dynamical information are well known. Generally, the fuzzy model is a correct zero-order representation of a process or function. However the derivative of its mathematical expression is not necessarily a derivative fuzzy model of the process or function. A perturbed fuzzy system, as a generalization of the traditional fuzzy system, is proposed. It has the ability to uniformly approximate continuous functions and their derivatives on arbitrarily compact sets to the desired degree.

In this study, we are concerned with a new category of logic connectives and logic neurons based on the concept of uninorms. Uninorms are a generalization of t -norms and t -conorms used for composing fuzzy sets. We discuss the development of such constructs by using genetic algorithms. In this way we optimize a suite of parameters encountered in uninorms, especially their identity element. In the sequel, we introduce a class of logic neurons based on uninorms (which will be refereed to as unineurons). The learning issues of the neurons are presented and some experimental results obtained for synthetic and benchmark data are reported.

In the resolution of group decision making problems where the consensus process can not be held face to face by the experts it is usually difficult for them to be able to identify the closeness of the opinions of the rest of the experts, and thus, it is difficult to have a clear view of the current state of the consensus process. In this paper we present a tool that creates consensus diagrams that can help experts to easily comprehend the current consensus state and to easily identify the experts that have similar or very different opinions. Those diagrams are based on several new similarity and consistency measures.