| Summary: | During the last years, artificial intelligence evolved rapidly by the introduction of some innovative approaches and theories such as fuzzy logic. Fuzzy systems have filled some important niches in several applications of control, decision, etc. Despite these successes, fuzzy logic has confronted rapidly its limitations by the lack of processing the effect of context. Human is capable to decide adequately fuzzy representations in particular contexts. Nevertheless a fuzzy system, once conceived by an expert, is non-adaptable to conditions that change under the effect of context. This problem has been addressed by several researchers; however, very few have touched it. In this research, it has been first identified the reasons of such an indifference in fuzzy reasoning. By complex implications at the cognitive level, several areas such as linguistics, cognitive psychology, artificial intelligence, the theory of knowledge and of memory and memorisation have been studied. These interdisciplinary studies helped to deduce essentials of such a model of contextual effect. Consequently, a cognitive structure named SFC is proposed. This structure contains internal mechanisms simulating human reasoning under contextual dependence. Proposed approach is structuralist by the fact of including internal modules, each one having different functionality and responsibilities. These modules interact hierarchically to process fuzzy information under the effect of context. As of representing knowledge, the approach assumes integrity at the level of information: knowledge to process has to contain, implicitly or explicitly, the information of context. Consequently, the effect of context is itself an information to process and to memorise. Under this viewpoint, knowledge is integrated to its context. The SFC structure is validated by an automatic conduct example of a car. Finally, a simplified structure named SFC-R that can be used in large-scale fuzzy systems is proposed. This structure is not only useful to reduce dimension in large scale rule bases, but also to integrate, in a very efficient manner, the discreet information in fuzzy rule bases.
|
|---|
