Techniques for the Fusion of Symbolic Rules in Distributed Organic Systems

Abstract

Humans do not only learn by their own experience but also by rules obtained from other humans. It is a challenging idea to enable distributed, intelligent computer systems to follow this human archetype. A basic technique needed for such an "organic" system is the fusion of functional knowledge in form of symbolic rules that are gained from several sources (nodes of the distributed system). We assume that these nodes are equipped with self-learning classifiers on the basis of a hybrid radial basis function network / fuzzy system paradigm. We provide methods for the fusion of fuzzy-type rules extracted from such classifiers. These methods aim at preserving the consistency and comprehensibility of a found rule set (e.g. low number of rules, distinguishability of membership functions) by means of a regularization approach.

Basic Technologies for Knowledge Transfer in Intelligent Systems

“…New training techniques are needed to enforce the interpretability and representativity of rules and classifiers. We must improve the techniques for rule integration: Similar rules (or linguistic terms) must be fused -we discuss a first approach in [3] to keep the classifiers interpretable. Also, we will develop mechanisms for an active measurement of other individuals' competence.…”

“…Our work has been inspired by various research areas: An appropriate machine learning paradigm, for example, could be found in the area of Soft Computing [2]. There is a few work on knowledge extraction in this field, too (see [3] for some references). Clustering methods could be adopted from the field of Pattern Recognition [4].…”

“…[3], [7]) as a hybrid system that can be seen as both, an RBF neural network (NN) and a Mamdani-type fuzzy system (FS). With the following definition we gain the advantages of two worlds: Trainability of NN and interpretability of FS (cf.…”

“…In the future, further improvements will be provided, e.g., techniques that keep the number of linguistic terms as low as possible. In [3] we describe how this can be done in principle. Also, there are no special methods for communication up to now; we simply broadcast the rules.…”

So near and yet so far: New insight into properties of some well-known classifier paradigms

“…For example, generative properties of RBF are improved by means of regularization terms in the objective function in [11]. The relevance vector machine [71] aims at improving the generative properties of SVM.…”