Narrow Results

A novel algorithm for unsupervised fuzzy clustering is introduced. The algorithm uses a so-called Weighted Fixed Neural Network (WFNN) to store important and useful information about the topological relations in a given data set. The algorithm produces a weighted connected net, of weighted nodes connected by weighted edges, which reflects and preserves the topology of the input data set. The weights of the nodes and the edges in the resulting net are proportional to the local densities of data samples in input space. The connectedness of the net can be changed, and the higher the connectedness of the net is chosen, the fuzzier the system becomes. The new algorithm is computationally efficient when compared to other existing methods for clustering multi-dimensional data, such as color images.

A new more efficient variant of a recently developed algorithm for unsupervised fuzzy clustering is introduced. A Weighted Incremental Neural Network (WINN) is introduced and used for this purpose. The new approach is called FC-WINN (Fuzzy Clustering using WINN). The WINN algorithm produces a net of nodes connected by edges, which reflects and preserves the topology of the input data set. Additional weights, which are proportional to the local densities in input space, are associated with the resulting nodes and edges to store useful information about the topological relations in the given input data set. A fuzziness factor, proportional to the connectedness of the net, is introduced in the system. A watershed-like procedure is used to cluster the resulting net. The number of the resulting clusters is determined by this procedure. Only two parameters must be chosen by the user for the FC-WINN algorithm to determine the resolution and the connectedness of the net. Other parameters that must be specified are those which are necessary for the used incremental neural network, which is a modified version of the Growing Neural Gas algorithm (GNG). The FC-WINN algorithm is computationally efficient when compared to other approaches for clustering large high-dimensional data sets.

The direct adaptive regulation of unknown nonlinear dynamical systems in Brunovsky form with modeling error effects, is considered in this paper. Since the plant is considered unknown, we propose its approximation by a special form of a Brunovsky type neuro–fuzzy dynamical system (NFDS) assuming also the existence of disturbance expressed as modeling error terms depending on both input and system states plus a not-necessarily-known constant value. The development is combined with a sensitivity analysis of the closed loop and provides a comprehensive and rigorous analysis of the stability properties. The existence and boundness of the control signal is always assured by introducing a novel method of parameter hopping and incorporating it in weight updating laws. Simulations illustrate the potency of the method and its applicability is tested on well known benchmarks, as well as in a bioreactor application. It is shown that the proposed approach is superior to the case of simple recurrent high order neural networks (HONN's).

A method frequently used in classification systems for improving classification accuracy is to combine outputs of several classifiers. Among various types of classifiers, fuzzy ones are tempting because of using intelligible fuzzy if-then rules. In the paper we build an AdaBoost ensemble of relational neuro-fuzzy classifiers. Relational fuzzy systems bond input and output fuzzy linguistic values by a binary relation; thus, fuzzy rules have additional, comparing to traditional fuzzy systems, weights — elements of a fuzzy relation matrix. Thanks to this the system is better adjustable to data during learning. In the paper an ensemble of relational fuzzy systems is proposed. The problem is that such an ensemble contains separate rule bases which cannot be directly merged. As systems are separate, we cannot treat fuzzy rules coming from different systems as rules from the same (single) system. In the paper, the problem is addressed by a novel design of fuzzy systems constituting the ensemble, resulting in normalization of individual rule bases during learning. The method described in the paper is tested on several known benchmarks and compared with other machine learning solutions from the literature.

In this paper we analyze the identification problem which consists of choosing an appropriate identification model and adjusting its parameters according to some adaptive law, such that the response of the model to an input signal (or a class of input signals), approximates the response of the real system for the same input. For identification models we use fuzzy-recurrent high order neural networks. High order networks are expansions of the first-order Hopfield and Cohen-Grossberg models that allow higher order interactions between neurons. The underlying fuzzy model is of Mamdani type assuming a standard defuzzification procedure such as the weighted average. Learning laws are proposed which ensure that the identification error converges to zero exponentially fast or to a residual set when a modeling error is applied. There are two core ideas in the proposed method: (1) Several high order neural networks are specialized to work around fuzzy centers, separating in this way the system into neuro-fuzzy subsystems, and (2) the use of a novel method called switching parameter hopping against the commonly used projection in order to restrict the weights and avoid drifting to infinity.

In this paper, some fundamental issues of modern multi-agent robot architectures are discussed. It is argued that the multi-agent approach provides the necessary flexibility and adaptivity for such architectures, and that the primary issue in designing a multi-agent robot architecture is the selection of the granularity level, i.e., the decision on decomposing the overall desired functionality physically or across tasks. It is explained why at the various system levels different decomposition grains are needed; physical components, tasks or hybrid. This granularity decision is made on the basis of specific criteria of control localization, knowledge decoupling and interaction minimization so as to identify the decision points of the overall functionality. The above criteria lead to a dual composition-decomposition relation, which provides a good basis for system scaling. The paper specializes the discussion to a proposed neuro-fuzzy multi-agent architecture, which is then applied to design the local path planning system of an indoor mobile robot.

Fault prognosis in industrial plants is a complex problem, and time is an important factor for the resolution of this problem. The main indicator for the task of fault prognosis is the estimate of remaining useful life (RUL), which essentially depends on the predicted time to failure. This paper introduces a temporal neuro-fuzzy system (TNFS) for performing the fault prognosis task and exactly estimating the RUL of preheater cyclones in a cement plant. The main component of the TNFS is a set of temporal fuzzy rules that have been chosen for their ability to explain the behavior of the entire system, the components’ degradation, and the RUL estimation. The benefit of introducing time in the structure of fuzzy rules is that a local memory of the TNFS is created to capture the dynamics of the prognostic task. More precisely, the paper emphasizes improving the performance of TNFSs for prediction. The RUL estimation process is broken down into four generic processes: building a predictive model, selecting the most critical parameters, training the TNFS, and predicting RUL through the generated temporal fuzzy rules. Finally, the performance of the proposed TNFS is evaluated using a real preheater cement cyclone dataset. The results show that our TNFS produces better results than classical neuro-fuzzy systems and neural networks.

Modeling and forecasting highly volatile time series is a complex task not suited for linear models. An example of such series is the electricity spot prices in Brazil. In this article we apply a hybrid neuro-fuzzy/neural network system to forecast the weekly spot in the Southeast region of Brazil up to six weeks in advance. The input variables used are lagged values of the Natural Inflow Energy (for the entire region and its two main basins) and of the spot prices themselves. The hybrid system starts by fitting a neural network to data; the result of this network is then saved and used as an additional input at an ANFIS type neuro-fuzzy structure. The forecasting performance of the proposed hybrid model is presented for different periods and the results discussed.

The paper presents an approach to classification based on neuro-fuzzy systems and hybrid learning algorithms. A new method of rule generation is proposed. The rules are used in order to create a connectionist neuro-fuzzy architecture of the multi-NF system. Parameters of the rules are then adjusted by a gradient algorithm. Thus, the system can be employed to solve multi-class classification problems. Some examples are depicted.

This chapter discusses some of the relations between CI (Computational Intelligence) and traditional, symbolic AI (Artificial Intelligence) systems. It also presents a comparative analysis of fuzzy, neural and genetic, as well as symbolic AI systems. This analysis allows us to determine the most promising directions in the synthesis of CI systems. Also, a concrete implementation of a CI system for decision support purposes (a neuro-fuzzy-genetic classifier) is presented. It is able to learn and generalize from learned knowledge as well as explain (by providing and optimizing a set of fuzzy rules describing the underlying mechanisms) the decisions it makes. Its application to the design of decision support system for the glass identification problem from the domain of forensic science is also presented. A comparative analysis of the proposed methodology with several other approaches has also been performed.