Adaptation of Connectionist Weighted Fuzzy Logic Programs with Kripke-Kleene Semantics

Authors:
Alexandros Chortaras;Giorgos Stamou;Andreas Stafylopatis;Stefanos Kollias
Affiliations:
School of Electrical and Computer Engineering, National Technical University of Athens, Athens, Greece;School of Electrical and Computer Engineering, National Technical University of Athens, Athens, Greece;School of Electrical and Computer Engineering, National Technical University of Athens, Athens, Greece;School of Electrical and Computer Engineering, National Technical University of Athens, Athens, Greece
Venue:
ICANN '08 Proceedings of the 18th international conference on Artificial Neural Networks, Part I
Year:
2008

Citing 5
Cited 0

Bilattices and the semantics of logic programming

Journal of Logic Programming
A Parametric Approach to Deductive Databases with Uncertainty

IEEE Transactions on Knowledge and Data Engineering
Algorithms of Nondifferentiable Optimization: Development and Application

Cybernetics and Systems Analysis
Nonsmooth training of fuzzy neural networks

Soft Computing - A Fusion of Foundations, Methodologies and Applications
Adaptation of weighted fuzzy programs

ICANN'06 Proceedings of the 16th international conference on Artificial Neural Networks - Volume Part II

Quantified Score

Hi-index	0.00

Visualization

Abstract

Weighted fuzzy logic programs extend the expressiveness of fuzzy logic programs by allowing the association of a different significance weight with each atom that appears in the body of a fuzzy rule. The semantics and a connectionist representation of these programs have already been studied in the absence of negation; in this paper we first propose a Kripke-Kleene based semantics for the programs which allows for the use of negation as failure. Taking advantage of the increased modelling capabilities of the extended programs, we then describe their connectionist representation and study the problem of adapting the rule weights in order to fit a provided dataset. The adaptation algorithm we develop is based on the subgradient descent method and hence is appropriate to be employed as a learning algorithm for the training of the connectionist representation of the programs.