Abstract not found

In this paper, we present a novel and efficient approach for automatic design of Artificial Neural Networks (ANNs) by evolving to the optimal network configuration(s) within an architecture space. The evolution technique, the so-called multidimensional Particle Swarm Optimization (MD PSO) re-forms the native structure of PSO particles in such a way that they can make inter-dimensional passes with a dedicated dimensional PSO process. So in a multidimensional search space where the optimum dimension is unknown, swarm particles can seek for both positional and dimensional optima. This eventually removes the necessity of setting a fixed dimension a priori, which is a common drawback for the family of swarm optimizers. With the proper encoding of the network configurations and parameters into particles, MD PSO can then seek for positional optimum in the error space and dimensional optimum in the architecture space. The optimum dimension converged at the end of a MD PSO process corresponds to a unique ANN configuration where the network parameters (connections, weights and biases) can then be resolved from the positional optimum reached on that dimension. The efficiency and performance of the proposed technique is demonstrated over one of the hardest synthetic problems. The experimental results show that MD PSO evolves to optimum or near-optimum networks in general. 1.

In current paper, we put forward a design method to Fault Diagnostic Systems#FDS# by proposing a new fault model and using the incremental hybrid learning algorithm which tightly combines symbolic learning and neural networks. Its capable of overcoming several shortcomings in the existing diagnostic systems, such as the lack of universality, the unbalance in the use of fault prior knowledge and the dynamic data and the dilemma of stability and plasticity. The experiment showed the FDS implemented by this kind of method had a good diagnostic ability.

In this paper, a fast adaptive neural regression estimator named FANRE is proposed. FANRE exploits the advantages of both Adaptive Resonance Theory and Field Theory while contraposing the Characteristic of regression problems. It achieves not only impressive approximating results but also fast learning speed. Besides, FANRE has incremental learning ability.

This paper examines a decision-tree framework for instance-space decomposition. According to the framework, the original instance-space is hierarchically partitioned into multiple subspaces and a distinct classifier is assigned to each subspace. Subsequently, an unlabeled, previously-unseen instance is classified by employing the classifier that was assigned to the subspace to which the instance belongs. After describing the framework, the paper suggests a novel splitting-rule for the framework and presents an experimental study, which was conducted, to compare various implementations of the framework. The study indicates that using the novel splitting-rule, previously presented implementations of the framework, can be improved in terms of accuracy and computation time.

Abstract. Neural network technology has already been applied in a variety of domains with remarkable success. However, it has not been well utilized in data mining and knowledge discovery. In this paper, a general neural framework named NEUCRUM is proposed for classiÞcation rule mining. This paper also presents a possible implementation of NEUCRUM whose key components are a speciÞc neural classiÞer named FANNC and a novel rule extraction approach named STARE.FANNC is used to learn from pre-processed data, in which its fast learning speed and strong generalization ability are quite contributive. STARE is proposed in this paper, which is used to extract comprehensible, compact and accurate symbolic rules from trained neural networks so that the knowledge discovered is explicitly available to decision-makers. Applications show that NEUCRUM and its implementation described in this paper work well in many real domains.

In this paper, we propose a methodology for the characterization of the suspended particulate matter along the Algiers’s bay. An approach by multi-layer perceptron (MLP) with training by back propagation of the gradient optimized by the algorithm of Levenberg-Marquardt (LM) is used. The accent was put on the choice of the components of the base of training where a comparative study made for four methods: Random and three alternatives of classification by K-Means. The samples are taken from suspended matter image, obtained by an analytical model based on polynomial regression by taking account of in situ measurements. The mask which selects the region of interest (water in our case) was done by using a multi spectral classification with ISODATA algorithm. To improve the result of classification, a cleaning of this mask was carried out using the mathematical morphology tools.

Abstract- An approach for invariant clustering and recognition of objects (situation) in dynamic environment is proposed. This approach is based on the combination of clustering by using unsupervised neural network (in particular ART-2) and preprocessing of sensor information by using forward multilayer perceptron (MLP) with error back propagation (EBP) which supervised by clustering neural network. Using MLP with EBP allows to recognize a pattern with relatively small transformations (shift, rotation, scaling) as a known previous cluster and to reduce producing large number of clusters in dynamic environment, e.g. during movement of robot or recognition of novelty in security system. I.