Neural Computing and Applications

In this paper, the method of the graphical interpretation of the single-layer network weights is introduced. It is shown that the network parameters can be converted to the image and their particular elements are the pixels. For this purpose, weight-to-pixel conversion formula is used. Moreover, new weights' modification method is proposed. The weight coefficients are computed on the basis of pixel values for which image filtration algorithms are implemented. The approach is applied to the weights of three types of the models: single-layer network, two-layer backpropagation network and the hybrid network. The performance of the models is then compared on two independent data sets. By means of the experiments, it is presented that the adjustment of the weights to new values decreases test error value compared to the error obtained for initial set of weights.

In this paper, we give characterizations of ordered semigroups in terms of (∈, ∈ ∨q)-fuzzy interior ideals. We characterize different classes regular (resp. intra-regular, simple and semisimple) ordered semigroups in terms of (∈, ∈ ∨q)-fuzzy interior ideals (resp. (∈, ∈ ∨q)-fuzzy ideals). In this regard, we prove that in regular (resp. intra-regular and semisimple) ordered semigroups the concept of (∈, ∈ ∨q)-fuzzy ideals and (∈, ∈ ∨q)-fuzzy interior ideals coincide. We prove that an ordered semigroup S is simple if and only if it is (∈, ∈ ∨q)-fuzzy simple. We characterize intra-regular (resp. semisimple) ordered semigroups in terms of (∈, ∈ ∨q)-fuzzy ideals (resp. (∈, ∈ ∨q)-fuzzy interior ideals). Finally, we consider the concept of implication-based fuzzy interior ideals in an ordered semigroup, in particular, the implication operators in Lukasiewicz system of continuous-valued logic are discussed.

In this paper, global exponential synchronization of a class of discrete delayed complex networks with switching topology has been investigated by using Lyapunov-Ruzimiki method. The impulsive scheme is designed to work at the time instant of switching occurrence. A time-varying delay-dependent criterion for impulsive synchronization is given to ensure the delayed discrete complex networks switching topology tending to a synchronous state. Furthermore, a numerical simulation is given to illustrate the effectiveness of main results.

Emotion recognition in speech is a topic on which little research has been done to date. In this paper, we discuss why emotion recognition in speech is an interesting and applicable research topic and present a system for emotion recognition using one-class-in-one neural networks. By using a large database of phoneme balanced words, our system is speaker and context independent. We achieve a recognition rate of approximately 50% when testing eight emotions

The process of reconstructing an original image from a compressed one is a difficult problem, since a large number of original images lead to the same compressed image and solutions to the inverse problem cannot be uniquely determined. Vector quantization is a compression technique that maps an input set of k-dimensional vectors into an output set of k-dimensional vectors, such that the selected output vector is closest to the input vector according to a selected distortion measure. In this paper, we show that adaptive 2D vector quantization of a fast discrete cosine transform of images using Kohonen neural networks outperforms other Kohonen vector quantizers in terms of quality (i.e. less distortion). A parallel implementation of the quantizer on a network of SUN Sparcstations is also presented.

This paper describes a new approach to the analysis of weather radar data for short-range rainfall forecasting based on a neural network model. This approach consists in extracting synthetic information from radar images using the approximation capabilities of multilayer neural networks. Each image in a sequence is approximated using a modified radial basis function network trained by a competitive mechanism. Prediction of the rain field evolution is performed by analysing and extrapolating the time series of weight values. This method has been compared to the conventional cross-correlation technique and the persistence method for three different rainfall events, showing significant improvement in 30 and 60 min ahead forecast accuracy.

A recent novel approach to the visualisation and analysis of datasets, and one which is particularly applicable to those of a high dimension, is discussed in the context of real applications. A feed-forward neural network is utilised to effect a topographic, structure-preserving, dimension-reducing transformation of the data, with an additional facility to incorporate different degrees of associated subjective information. The properties of this transformation are illustrated on synthetic and real datasets, including the 1992 UK Research Assessment Exercise for funding in higher education. The method is compared and contrasted to established techniques for feature extraction, and related to topographic mappings, the Sammon projection and the statistical field of multidimensional scaling.

The abrasion resistance of chenille yarn is crucially important in particular because the effect sought is always that of the velvety feel of the pile. Thus, various methods have been developed to predict chenille yarn and fabric abrasion properties. Statistical models yielded reasonably good abrasion resistance predictions. However, there is a lack of study that encompasses the scope for predicting the chenille yarn abrasion resistance with artificial neural network (ANN) models. This paper presents an intelligent modeling methodology based on ANNs for predicting the abrasion resistance of chenille yarns and fabrics. Constituent chenille yarn parameters like yarn count, pile length, twist level and pile yarn material type are used as inputs to the model. The intelligent method is based on a special kind of ANN, which uses radial basis functions as activation functions. The predictive power of the ANN model is compared with different statistical models. It is shown that the intelligent model improves prediction performance with respect to statistical models.

One of the major tasks in some human–computer interface applications, such as face recognition and video telephony, is to localize a human face in an image. In this paper, we propose to use hierarchical neural networks with local recurrent connectivity to solve this task not only in unambiguous situations, but also in the presence of complex backgrounds, difficult lighting, and noise. The networks are trained using a database of gray-scale still images and manually determined eye coordinates. They are able to produce reliable and accurate eye coordinates for unknown images by iteratively refining initial solutions. Because the networks process entire images, there is no need for any time-consuming scanning across positions and scales. Furthermore, the fast network updates allow for real-time face tracking. In this case, the networks are trained using still images that move in random directions. The trained networks are able to accurately track the eye positions in the test image sequences.

This work proposes decomposition of square approximation algorithm for neural network weights update. Suggested improvement results in alternative method that converge in less iteration and is inherently parallel. Decomposition enables parallel execution convenient for implementation on computer grid. Improvements are reflected in accelerated learning rate which may be essential for time critical decision processes. Proposed solution is tested and verified on multilayer perceptrons neural network case study, varying a wide range of parameters, such as number of inputs/outputs, length of input/output data, number of neurons and layers. Experimental results show time savings up to 40% in multiple thread execution. KeywordsNeural network-Weights update-Gradient learning method-Parallel processing

In this study, the traffic accidents recognizing risk factors related to the environmental (climatological) conditions that are associated with motor vehicles accidents on the Konya-Afyonkarahisar highway with the aid of Geographical Information Systems (GIS) have been determined using the combination of K-means clustering (KMC)-based attribute weighting (KMCAW) and classifier algorithms including artificial neural network (ANN) and adaptive network-based fuzzy inference system (ANFIS). The dynamic segmentation process in ArcGIS9.0 from the traffic accident reports recorded by District Traffic Agency has identified the locations of the motor vehicle accidents. The attributes obtained from this system are day, temperature, humidity, weather conditions, and month of occurred traffic accidents. The traffic accident dataset comprises five attributes (day, temperature, humidity, weather conditions, and month of occurred traffic accidents) and 358 observations including 179 without accident and 179 with accident. The proposed comprises two stages. In the first stage, the all attributes of dataset have been weighted using KMCAW method. The aims of this weighting method are both to increase the classification performance of used classifier algorithm and to transform from linearly non-separable traffic accidents dataset to a linearly separable dataset. In the second stage, after weighting process, ANN and ANFIS classifier algorithms have been separately used to determine the case of traffic accidents as with accident or without accident. In order to evaluate the performance of proposed method, the classification accuracy, sensitivity, specificity and area under the ROC (Receiver Operating Characteristic) curves (AUC) values have been used. While ANN and ANFIS classifiers obtained the overall prediction accuracies of 53.93 and 38.76%, respectively, the combination of KMCAW and ANN and the combination of KMCAW and ANFIS achieved the overall prediction accuracies of 74.15 and 55.06% on the prediction of traffic accidents. The experimental results have demonstrated that the proposed attribute weighting method called KMCAW is a robust and effective data pre-processing method in the prediction of traffic accidents on Konya-Afyonkarahisar highway in Turkey. KeywordsGeographical information systems (GIS)–Traffic accident analysis–Prediction–K-means clustering based attribute weighting–Artificial neural network–Adaptive network based fuzzy inference system

The ability of neural networks to learn from repeated exposure to system characteristics has made them a popular choice for many applications in linear and non-linear control. In this paper, the capabilities of neural networks in detecting and accommodating control surface failures for a modified F/A-18 super-manoeuverable fighter aircraft are examined. To detect and accommodate a failure in the thrust vectoring vane during a pitch manoeuvre, a hierarchical neuro-controller is designed using thrust vectoring, symmetric leading edge flap and the throttle. This neuro- controller is then used as the fault accommodating neuro- controller. A separate neural network is trained to detect failures in the thrust vectoring vane. The performance of the controller and fault-detection networks are verified using a numerical simulation of a longitudinal model of the aircraft.

Artificial Neural Networks (ANNs) have recently become the focus of considerable attention in many disciplines, including robot control, where they can be used as a general class of nonlinear models to solve highly nonlinear control problems. Feedforward neural networks have been widely applied for modelling and control purposes. One of the ANN applications in robot control is for the solution of the inverse kinematic problem, which is important in path planning of robot manipulators. This paper proposes an iterative approach and an offset error compensation method to improve the accuracy of the inverse kinematic solutions by using an ANN and a forward kinematic model of a robot. The offset error compensation method offers potential to generate accurately the inverse solution for a class of problems which have an easily obtained forward model and a complicated solution.

The HONEST neural network is a recently-developed generalisation of the well-known Sigma-Pi high order neural network. HONEST has previously been applied to diabetes forecasting and feature combination in an Othello evaluation function. In this paper, we apply HONEST to the classification into age-groups of abalone shellfish, a difficult bench-mark to which previous researchers have applied cascade correlation, standard backpropagation with a Multi-Layer Perceptron (MLP) network, Quinlan's C4.5, and the DYSTAL network. While the best reported test set performance by previous researchers is 65.61% correct classification, HONEST was able to achieve 72.89% correct test set classification. In addition, HONEST's transparent structure allows us to manually examine the network state and make observations about the solution the network has learned.

Due to deregulation of electricity industry, accurate load forecasting and predicting the future electricity demand play an important role in the regional and national power system strategy management. Electricity load forecasting is a challenging task because electric load has complex and nonlinear relationships with several factors. In this paper, two hybrid models are developed for short-term load forecasting (STLF). These models use “ant colony optimization (ACO)” and “combination of genetic algorithm (GA) and ACO (GA-ACO)” for feature selection and multi-layer perceptron (MLP) for hourly load prediction. Weather and climatic conditions, month, season, day of the week, and time of the day are considered as load-influencing factors in this study. Using load time-series of a regional power system, the performance of ACO+MLP and GA-ACO+MLP hybrid models is compared with principal component analysis (PCA)+MLP hybrid model and also with the case of no-feature selection (NFS) when using MLP and radial basis function (RBF) neural models. Experimental results and the performance comparison with similar recent researches in this field show that the proposed GA-ACO+MLP hybrid model performs better in load prediction of 24-h ahead in terms of mean absolute percentage error (MAPE). KeywordsShort-term load forecasting–Feature selection–Ant colony optimization–Genetic algorithm–Neural network

This paper describes a novel approach to the simulation of language disorders, based upon the notion of a multi-network architecture —a set of autonomous neural networks which have been linked in some manner to perform a complex function that cannot readily be performed by any one network alone. The merits of this approach have been assessed by mapping a neuropsychological model of single-word language processing onto a multi-network architecture. Language disorders may be simulated by damaging, or lesioning, one or more component networks. Our attempts to simulate two specific language disorders, semantic dementia and deep dysphasia, are described. The relative success of our simulation work is encouraging, and leads us to conclude that a multi-network approach to the simulation of cognitive function and dysfunction offers a valid alternative to the traditional single-network based perspective.

This paper presents an automated knowledge acquisition architecture for the truck docking problem. The architecture consists of a neural network block, a fuzzy rule generation block and a genetic optimisation block. The neural network block is used to quickly and adaptively learn from trials the driving knowledge. The fuzzy rule generation block then extracts the driving knowledge to form a knowledge rule base. The driving knowledge rule base is further optimised in the genetic optimisation block using a genetic algorithm. Computer simulations are presented to show the effectiveness of the architecture.

Neural network technology is experiencing rapid growth and is receiving considerable attention from almost every field of science and engineering. The attraction is due to the successful application of neural network techniques to several real world problems. Neural networks have not yet found widespread application in weather forecasting. The reason for this has been the difficulty in obtaining suitable weather forecasting data sets. In this paper we describe our experience in applying neural network techniques for acquiring the necessary knowledge to predict the weather conditions of Melbourne City and its suburbs in Australia during a 24 hour period beginning at 9 am local time. The accuracy of forecasts produced by a given forecasting procedure typically varies with factors such as geographical location, season, categories of weather, quality of input data, lead time and validity time. Two types of weather data sets assembled from the archives of the Australian Commonwealth Bureau of Meteorology are used for training the neural network. The results of the experiments are competitive and are discussed.

In artificial neural networks (ANNs), the activation function most used in practice are the logistic sigmoid function and the hyperbolic tangent function. The activation functions used in ANNs have been said to play an important role in the convergence of the learning algorithms. In this paper, we evaluate the use of different activation functions and suggest the use of three new simple functions, complementary log-log, probit and log-log, as activation functions in order to improve the performance of neural networks. Financial time series were used to evaluate the performance of ANNs models using these new activation functions and to compare their performance with some activation functions existing in the literature. This evaluation is performed through two learning algorithms: conjugate gradient backpropagation with Fletcher–Reeves updates and Levenberg–Marquardt. KeywordsNeural networks–Activation functions–Complementary log-log–Probit–Log-log–CGF algorithm–LM algorithm

This paper describes the visual feedback positional control of the XY piezo actuator stage (PAS). The XY PAS control system consists of four main components, i.e. XY PAS as controlled object, supply electronics for piezoelectric actuators (PEAs), microscope with digital camera for visualization and for measuring the actual position and a vision processing module in combination with a desktop PC, as processing hardware. XY PAS is fabricated by a photo structuring process from photosensitive glass, and PEAs are built-onto meet the request for its precise movement. It is evident from the electromechanical model of XY PAS, that accurate positioning of XY PAS is an exacting piece of work, due to the nonlinear hysteresis inherent in PEAs. Accordingly, two neural network control techniques were developed, i.e. the feedforward neural network controller (FFNNC) and the feedforward/feedback neural network controller (FF/FBNNC). Proposed neural network controllers are compared with the traditional linear controllers. KeywordsPiezo actuator stage-Position control-Neural networks-Nonlinear hysteresis

Due to mobility of wireless hosts, routing in mobile ad-hoc networks (MANETs) is a challenging task. Multipath routing is employed to provide reliable communication, load balancing, and improving quality of service of MANETs. Multiple paths are selected to be node-disjoint or link-disjoint to improve transmission reliability. However, selecting an optimal disjoint multipath set is an NP-complete problem. Neural networks are powerful tools for a wide variety of combinatorial optimization problems. In this study, a transient chaotic neural network (TCNN) is presented as multipath routing algorithm in MANETs. Each node in the network can be equipped with a neural network, and all the network nodes can be trained and used to obtain optimal or sub-optimal high reliable disjoint paths. This algorithm can find both node-disjoint and link-disjoint paths with no extra overhead. The simulation results show that the proposed method can find the high reliable disjoint path set in MANETs. In this paper, the performance of the proposed algorithm is compared to the shortest path algorithm, disjoint path set selection protocol algorithm, and Hopfield neural network (HNN)-based model. Experimental results show that the disjoint path set reliability of the proposed algorithm is up to 4.5times more than the shortest path reliability. Also, the proposed algorithm has better performance in both reliability and the number of paths and shows up to 56% improvement in path set reliability and up to 20% improvement in the number of paths in the path set. The proposed TCNN-based algorithm also selects more reliable paths as compared to HNN-based algorithm in less number of iterations. KeywordsTransient chaotic neural network–Mobile ad-hoc network–Disjoint multipath routing–Reliability

We propose a learning method for the ADALINE. The proposed method exploits fuzzy logic system for automatic tuning of the weights of the ADALINE. The inputs of the fuzzy logic system are error and change of error, and the output is the weight variation. We used same membership functions and different scaling factor for each weights. In order to verify the effectiveness of the proposed method, we performed the simulation and experimentation for the cases of the noise cancellation and the inverted pendulum control. The results show that the proposed method does not need the learning rate and the derivative, and improves the performance compared to the Widrow–Hoff delta rule for ADALINE.

In this paper, we propose the problem of online cost-sensitive clas- sifier adaptation and the first algorithm to solve it. We assume we have a base classifier for a cost-sensitive classification problem, but it is trained with respect to a cost setting different to the desired one. Moreover, we also have some training data samples streaming to the algorithm one by one. The prob- lem is to adapt the given base classifier to the desired cost setting using the steaming training samples online. To solve this problem, we propose to learn a new classifier by adding an adaptation function to the base classifier, and update the adaptation function parameter according to the streaming data samples. Given a input data sample and the cost of misclassifying it, we up- date the adaptation function parameter by minimizing cost weighted hinge loss and respecting previous learned parameter simultaneously. The proposed algorithm is compared to both online and off-line cost-sensitive algorithms on two cost-sensitive classification problems, and the experiments show that it not only outperforms them one classification performances, but also requires significantly less running time.

This paper describes the use of an evolutionary design system known as GANNET to synthesize the structure of neural networks. Initial results are presented for two benchmark problems: the exclusive-or and the two-spirals. A variety of performance criteria and design components are used and comparisons are drawn between the performance of genetic algorithms and other related techniques on these problems.

This paper introduces a novel hybrid algorithm to determine the parameters of radial basis function neural networks (number of neurons, centers, width and weights) automatically. The hybrid algorithm combines the mix encoding particle swarm optimization algorithm with the back propagation (BP) algorithm to form a hybrid learning algorithm (MPSO-BP) for training Radial Basis Function Networks (RBFNs), which adapts to the network structure and updates its weights by choosing a special fitness function. The proposed method is used to deal with three nonlinear problems, and the results obtained are compared with existent bibliography, showing an improvement over the published methods.

In this paper, an approach to weighting features for classification based on the nearest-neighbour rules is proposed. The weights are adaptive in the sense that the weight values are different in various regions of the feature space. The values of the weights are found by performing a random search in the weight space. A correct classification rate is the criterion maximised during the search. Experimentally, we have shown that the proposed approach is useful for classification. The weight values obtained during the experiments show that the importance of features may be different in different regions of the feature space.

In this paper, an adaptive robust control scheme is developed which is suitable for the control of a class of uncertain nonlinear systems, typical of many servo manipulators. The control scheme is comprised of a model reference adaptive controller (MRAC) augmented with a nonlinear compensator based on an adaptive radial basis function (RBF). The RBF compensator is used to neutralise the effects of uncertain and possibly nonlinear dynamics, so that the equivalent system as seen by the MRAC is reduced to one without significant unstructured modelling errors. A stability analysis is provided to show the uniform stability and the asymptotic tracking capabilities of the proposed control system. Real-time experiment results verify the effectiveness of the control scheme.

In this paper, a recurrent neural network (RNN) based robust tracking controller is designed for a class of multiple-input-multiple-output discrete time nonlinear systems. The RNN is used in the closed-loop system to estimate online unknown nonlinear system function. A multivariable robust adaptive gradient-descent training algorithm is developed to train RNN. The weight convergence and system stability are proven in the sense of Lyapunov function. Simulation results are presented for a two-link robot tracking control problem. KeywordsRecurrent neural networks (RNNs)–Multivariable robust adaptive gradient-descent training algorithm (MRAGD)–Multiple-input-multiple-output (MIMO)–Stability

The structure of the extended fuzzy basis function network (EFBFN) is firstly proposed, and the least squares (LS) method is used to design it by fixing the widths of the hidden units in EFBFN. Then, to enhance the performance of the obtained EFBFN ulteriorly, a novel evolutionary algorithm based on LS and the hybrid of evolutionary programming and particle swarm optimization (LS-EPPSO) is proposed, in which we use EPPSO to tune the parameters of the premise part in EFBFN, and the LS algorithm to decide the consequent parameters in it simultaneously. The enhanced EFBFN whose parameters are refined automatically using LS-EPPSO is thus called adaptive EFBFN. In the simulation part, the proposed method to construct AEFBFN is employed to model a three input nonlinear function and to predict a chaotic time series. Comparisons with some typical fuzzy modeling methods and artificial neural networks are presented and discussed.

Radial basis function network (RBFN), commonly used in the classification applications, has two parameters, kernel center and radius that can be determined by unsupervised or supervised learning. But it has a disadvantage that it considers that all the independent variables have the equal weights. In that case, the contour lines of the kernel function are circular, but in fact, the influence of each independent variable on the model is so different that it is more reasonable if the contour lines are oval. To overcome this disadvantage, this paper presents an adaptive radial basis function network (ARBFN) with kernel shape parameters and derives the learning rules from supervised learning. To verify that this architecture is superior to that of the traditional RBFN, we make a comparison between three artificial and fifteen real examples in this study. The results show that ARBFN is much more accurate than the traditional RBFN, illustrating that the shape parameters can actually improve the accuracy of RBFN. KeywordsRadial basis function network–Supervised learning–Kernel function–Classification

A neural network architecture is introduced which implements a supervised clustering algorithm for the classification of feature vectors. The network is selforganising, and is able to adapt to the shape of the underlying pattern distribution as well as detect novel input vectors during training. It is also capable of determining the relative importance of the feature components for classification. The architecture is a hybrid of supervised and unsupervised networks, and combines the strengths of three wellknown architectures: learning vector quantisation, backpro-pagation and adaptive resonance theory. Network performance is compared to that of learning vector quantisation, back-propagation and cascade-correlation. It is found that performance is generally as good as or better than the performance of these other architectures, while training time is considerably shorter. However, the main advantage of the hybrid architecture is its ability to gain insight into the feature pattern space.

This paper considers the equalisation problem in Quadrature Phase-Shift Keying (QPSK) modulated signals which have been distorted by the passage through a transmission channel. The channel is modelled as a Rician fading channel to simulate the behaviour of the transmission channel in the mobile satellite context. The equalisation is treated as the generalisation of the channel behaviour, and some algorithms with the structure of an artificial neural network using the Multilayer Perceptron, Volterra Series and Radial Basis Function are described. Results for the BER performance of typical transversal equalisers, with Square-Root Kalman adaptation algorithm, and algorithms with artificial neural network structure are also reported and evaluated. Improved performance is exhibited by the artificial neural network approaches.

An algorithm for fast minimum search is proposed, which achieves very satisfying performance harmonising the Vogl's and the Conjugate Gradient algorithms. Such effectiveness is achieved by making adaptive, in a very simple and satisfactory way, both the learning rate and the momentum term, and by executing controls and corrections both on the possible cost function increase and on moves opposite to the direction of the negative of the gradient. Thanks to these improvements, we can obtain a good scaling relationship in learning. As regards the real world context, a musical application showed favourable results: besides the good convergence speed, a high generalisation capability has been achieved, as confirmed both by subjective musical evaluations and by objective tests.

Surface- and prototype-based models are often regarded as alternative paradigms to represent internal knowledge in trained neural networks. This paper analyses a network model (Circular Back-Propagation) that overcomes such dualism by choosing the best-fitting representation adaptively. The model involves a straightforward modification to classical feed-forward structures to let neurons implement hyperspherical boundaries; as a result, it exhibits a notable representation power, and benefits from the simplicity and effectiveness of classical back-propagation training. Artificial testbeds support the model definition by demonstrating its basic properties; an application to a real, complex problem in the clinical field shows the practical advantages of the approach.

The purpose of this paper is to design an adaptive controller and system experimental implementation for nonlinear translational oscillations with a rotational actuator (TORA) system. A wavelet-based neural network (WNN) is proposed to develop an adaptive backstepping control scheme, called ABCWNN for TORA system. To ensure the stability of the controlled system, a compensated controller is designed to enhance the control performance. Based on its universal approximation ability, we use a WNN to estimate the system uncertainty including frictional forces, external disturbance, and parameter variance. According to the estimations of the WNNs, the ABCWNN control is developed via a backstepping design procedure such that the system outputs follow the desired trajectories. For system development, the effects of frictional forces are discussed and solved using the estimation of the WNN. The effectiveness of the proposed control scheme for TORA system is verified by numerical simulation and experimental results. KeywordsBackstepping–Nonlinear control–Adaptive–Lyapunov theorem–Wavelet neural network

An RBF neural network-based adaptive control is proposed for Single-Input and Single-Output (SISO) linearisable nonlinear systems in this paper. It is shown that a SISO nonlinear system is first linearised by using the differential geometric approach in the state space, and the linearised nonlinear system is then treated as a partially known system. The known dynamics are used to design a nominal feedback controller to stabilise the nominal system, and an adaptive RBF neural network-based compensator is then designed to compensate for the effects of uncertain dynamics. The main function of the RBF neural network in this work is to adaptively learn the upper bound of the system uncertainty, and the output of the neural network is then used to adaptively adjust the gain of the compensator so that the strong robustness with respect to unknown dynamics can be obtained, and the tracking error between the plant output and the desired reference signal can asymptotically converge to zero. A simulation example is performed in support of the proposed scheme.

A plastic algorithm for building vector quantisers adaptively attains a dynamic representation of observed data; an unsupervised version of classical crossvalidation rules the algorithm's stopping condition. Combining plasticity with empirical generalisation-based control yields an adaptive methodology for VQ. The paper analyses the method's convergence properties and discusses the model's generalisation performance. Experimental results on synthetic and real, complex testbeds support the model's validity.

This paper describes a novel method of facial representation and recognition based upon adaptive processing of tree structures. Instead of the conventional flat vector representation for a face, a neural network approach-based technique is proposed to transform the Localised Gabor Feature (LGF) vectors extracted from human facial components into Human Face Tree Structure (HFTS) to represent a human face. A structural training algorithm is assigned to train and recognize the face identity in this HFTS representation with the corresponding LGF vectors. By benchmarking using the tested public face databases presented in this paper, our approach is able to achieve accuracy up to 90% under different scenarios of lighting conditions and posture orientations.

An adaptive learning algorithm is proposed for a feedforward neural network. The design principle is based on the sliding mode concept. Unlike the existing algorithms, the adaptive learning algorithm developed does not require a prioriknowledge of upper bounds of bounded signals. The convergence of the algorithm is established and conditions given. Simulations are presented to show the effectiveness of the algorithm.

In this paper, Bayesian network (BN) and ant colony optimization (ACO) techniques are combined in order to find the best path through a graph representing all available itineraries to acquire a professional competence. The combination of these methods allows us to design a dynamic learning path, useful in a rapidly changing world. One of the most important advances in this work, apart from the variable amount of pheromones, is the automatic processing of the learning graph. This processing is carried out by the learning management system and helps towards understanding the learning process as a competence-oriented itinerary instead of a stand-alone course. The amount of pheromones is calculated by taking into account the results acquired in the last completed course in relation to the minimum score required and by feeding this into the learning tree in order to obtain a relative impact on the path taken by the student. A BN is used to predict the probability of success, by taking historical data and student profiles into account. Usually, these profiles are defined beforehand; however, in our approach, some characteristics of these profiles, such as the level of knowledge, are classified automatically through supervised and/or unsupervised learning. By using ACO and BN, a fitness function, responsible for automatically selecting the next course in the learning graph, is defined. This is done by generating a path which maximizes the probability of each user’s success on the course. Therefore, the path can change in order to adapt itself to learners’ preferences and needs, by taking into account the pedagogical weight of each learning unit and the social behaviour of the system. KeywordsAnt colony optimization–Bayesian networks–Features modelling–OSGi–E-learning

Conventional adaptive control techniques have, for the most part, been based on methods for linear or weakly non-linear systems. More recently, neural network and genetic algorithm controllers have started to be applied to complex, non-linear dynamic systems. The control of chaotic dynamic systems poses a series of especially challenging problems. In this paper, an adaptive control architecture using neural networks and genetic algorithms is applied to a complex, highly nonlinear, chaotic dynamic system: the adaptive attitude control problem (for a satellite), in the presence of large, external forces (which left to themselves led the system into a chaotic motion). In contrast to the OGY method, which uses small control adjustments to stabilize a chaotic system in an otherwise unstable but natural periodic orbit of the system, the neuro-genetic controller may use large control adjustments and proves capable of effectively attaining any specified system state, with no a prioriknowledge of the dynamics, even in the presence of significant noise.

In this paper, we present a wavelet network IIR filtering system satisfying asymptotic stability in the sense of Lyapunov unlike many other gradient descent algorithms based adaptive filtering systems. The proposed system also carries the advantages of the time-frequency specific properties of wavelet networks embedded into the proposed filter dynamics. Two experiments for system identification problems corresponding to the infinite impulse response filter design are proposed. The results verified that the proposed wavelet network infinite impulse response adaptive filtering system not only performs better than gradient descent based algorithms but also performs as good as other stability theory based optimization algorithms.

Historical data for hospital admissions and Emergency Department (ED) visits in Baltimore City contain information concerning temporal patterns of paediatric asthma service utilisation (e.g. number of peaks and troughs, timing, relative magnitudes, steepness of rise and fall of the endemic cycles, etc.). This historical information can be captured by linear and neural network models to accurately predict the level of asthma admissions for the next few days or one week. Using 14 years of data, the best neural network models explained over 80% of the variations in admissions data with root mean square errors of 5–7 admissions per week. Models developed to predict asthma admissions can aid in identifying future peak periods of asthma admissions, alerting and educating individual asthmatic patients to periods of increased risk, and mitigating asthma events that lead to ED and/or hospital admissions. It is believed that these modelling techniques using historical data can be applied to any city or region with similar accuracies.

This work illustrates the use of neural networks for system identification of the dynamics of a distributed parameter system, an adsorption column for waste-water treatment of water containing toxic chemicals. System identification of this process is done from simulated data for this work. The inputs to the networks include the state of the column at a given point in time and the system input, the velocity. The network predicts the change in the state over a period of time based on these inputs. Recurrent networks were found to be capable of simulating the whole operation of the column from an initial state of zero concentrations throughout the column, and thus predicting the complete breakthrough curves. The feasibility of system identification of this process has been established using synthetic noisy data, which indicates that the same can be performed from experimental data when all the required measurements are available.

Advanced Manufacturing Technology (AMT) adoption can be complex, costly, and risky. Companies need to assess and evaluate their current conditions with that of AMT requirements to identify the gaps and predict their performance. Such an approach will facilitate companies not only in their investment decisions, but on the actions needed to improve performance. The lack of such an approach prompted this study to develop an Artificial Neural Network (ANN) classification and prediction model that can assist companies especially Small and Medium size Enterprises (SMEs) in evaluating AMT implementation. Data were collected from a survey of 140 SMEs. Using cluster analysis, the companies were classified into three groups based on their performance. Then, a feed-forward NN was developed and trained with back-propagation algorithm. The results showed that the model can classify companies with 72% accuracy rate into the three clusters. This model is suitable to evaluate AMTs implementation outcomes and predict company performance as high, low, or poor in technology adoption. KeywordsArtificial neural network–Advanced manufacturing technology–Small and medium size enterprises–Performance–Classification

This article proposes a reinforcement learning procedure for mobile robot navigation using a latent-like learning schema. Latent learning refers to learning that occurs in the absence of reinforcement signals and is not apparent until reinforcement is introduced. This concept considers that part of a task can be learned before the agent receives any indication of how to perform such a task. In the proposed topological reinforcement learning agent (TRLA), a topological map is used to perform the latent learning. The propagation of the reinforcement signal throughout the topological neighborhoods of the map permits the estimation of a value function which takes in average less trials and with less updatings per trial than six of the main temporal difference reinforcement learning algorithms: Q-learning, SARSA, Q()-learning, SARSA(), Dyna-Q and fast Q()-learning. The RL agents were tested in four different environments designed to consider a growing level of complexity in accomplishing navigation tasks. The tests suggested that the TRLA chooses shorter trajectories (in the number of steps) and/or requires less value function updatings in each trial than the other six reinforcement learning (RL) algorithms.

This paper proposes a pursuit system that utilizes the artificial life concept where autonomous mobile agents emulate the social behavior of animals and insects and realize their group behavior. Each agent contains sensors to perceive other agents in several directions, and decides its behavior based on the information obtained by these sensors. In this paper, a neural network is used for behavior decision controlling. The input of the neural network is decided by the existence of other agents, and the distance to the other agents. The output determines the directions in which the agent moves. The connection weight values of this neural network are encoded as genes, and the fitness individuals are determined using a genetic algorithm. Here, the fitness values imply how much group behavior adequately fit the goal and can express group behavior. The validity of the system is verified through simulation. Also in this paper, we have observed the agents emergent behavior during simulation.

Noisy and large data sets are extremely difficult to handle and especially to predict. Time series prediction is a problem, which is frequently addressed by researchers in many engineering fields. This paper presents a hybrid approach to handle a large and noisy data set. In fact, a Self Organizing Map (SOM), combined with multiple recurrent neural networks (RNN) has been trained to predict the components of noisy and large data set. The SOM has been developed to construct incrementally a set of clusters. Each cluster has been represented by a subset of data used to train a recurrent neural network. The back propagation through time has been deployed to train the set of recurrent neural networks. To show the performances of the proposed approach, a problem of instruction addresses prefetching has been treated.

The forecasting of air pollution is important for living environment and public health. The prediction of SO2 (sulfur dioxide), which is one of the indicators of air pollution, is a significant part of steps to be done in order to decrease the air pollution. In this study, a novel feature scaling method called neighbor-based feature scaling (NBFS) has been proposed and combined with artificial neural network (ANN) and adaptive network–based fuzzy inference system (ANFIS) prediction algorithms in order to predict the SO2 concentration value is from air quality metrics belonging to Konya province in Turkey. This work consists of two stages. In the first stage, SO2 concentration dataset has been scaled using neighbor-based feature scaling. In the second stage, ANN and ANFIS prediction algorithms have been used to forecast the SO2 value of scaled SO2 concentration dataset. SO2 concentration dataset was obtained from Air Quality Statistics database of Turkish Statistical Institute. To constitute dataset, the mean values belonging to seasons of winter period have been used with the aim of watching the air pollution changes between dates of December, 1, 2003 and December, 30, 2005. In order to evaluate the performance of the proposed method, the performance measures including mean absolute error (MAE), mean square error (MSE), root mean square error (RMSE), and IA (Index of Agreement) values have been used. After NBFS method applied to SO2 concentration dataset, the obtained RMSE and IA values are 83.87–0.27 (IA) and 93–0.33 (IA) using ANN and ANFIS, respectively. Without NBFS, the obtained RMSE and IA values are 85.31–0.25 (IA) and 117.71–0.29 (IA) using ANN and ANFIS, respectively. The obtained results have demonstrated that the proposed feature scaling method has been obtained very promising results in the prediction of SO2 concentration values. KeywordsNeighbor-based feature scaling–SO2 prediction–Feature scaling–Air pollution

In this paper, an easy and efficient method is brought forward to design the feedback control for the synchronization of two multiple time-delayed chaotic Hopfield neural networks, whose activation functions and delayed activation functions can have different forms of mapping. Without many complex restrictions and Lyapunov analytic process, the feedback control is given based on the M-matrix theory, the system parameters and the feedback section coefficients. All the results are simulated by Matlab and Simulink, which shows the simplicity and validity of the control. As shown in the simulation results, the error systems converge to zero rapidly. KeywordsSynchronization-Chaos-Hopfield neural network-Time delays-Algebraic condition