Advances in Neural Networks – ISNN 2007

In this paper, an adaptive fuzzy-neural network control problem is discussed for some uncertain MIMO nonlinear systemswith the block-triangular structure. The fuzzy-neural networks are utilized to approximate the virtual controllers, and by using backstepping technique, the direct adaptive FNN control scheme is developed. The proposed control method guarantees the closed-loop signals to be semiglobally uniformly ultimately bounded.

A newly developed non-symmetric sinusoidal membership function (NSSMF) is constructed. An improved fuzzy neural network controller using NSSMF is constructed to control the speed of ultrasonic motors. A dynamic algorithm with adaptive learning rate is used to train FNNC online. The global convergence of the FNNC systems could be guaranteed by adjusting the adaptive learning rate. The validity of the proposed scheme is examined by simulated experiments.

This paper proposes ANFIS logic based autonomous flight controller for UAVs (unmanned aerial vehicles). Three fuzzy logic modules are developed for the control of the altitude, the speed, and the roll angle, through which the altitude and the latitude-longitude of the air vehicle is controlled. The implementation framework utilizes MATLAB’s standard configuration and the Aerosim Aeronautical Simulation Block Set which provides a complete set of tools for rapid development of detailed 6 degree-of-freedom nonlinear generic manned/unmanned aerial vehicle models. The Aerosonde UAV model is used in the simulations in order to demonstrate the performance and the potential of the controllers. Additionally, Microsoft Flight Simulator and FlightGear Flight Simulator are deployed in order to get visual outputs that aid the designer in the evaluation of the controllers. Despite the simple design procedure, the simulated test flights indicate the capability of the approach in achieving the desired performance.

The special scenario of the topology in the cellular Ad Hoc networks was analyzed and a novel cross layer power control game algorithm based on Neural Fuzzy Connection Admission Controller (NFCAC) was proposed in this paper. NFCAC has been successfully applied in the control-related problems of neural networks. However, there is no discussion about the power control game algorithm and location recognition based on NFCAC in cellular Ad Hoc networks. The proposed algorithm integrated the attributes both of NFCAC and the topology space in special scenario. The topology and the power consumption of each node can all be optimized due to the minimum link occupation with the help of the algorithm. Simulation results show that the novel algorithm can give more power control guarantee to cellular Ad Hoc networks in the variable node loads and transmitting powers, and make the node more stable to support multi-hops at the same time.

This paper proposes a model predictive control scheme with recurrent fuzzy neural network (RFNN) by using the temperature of the drying process for grain dryers. In this scheme, there are two RFNNs and two PI controllers. One RFNN with feedforeward and feedback connections of grain layer history position states predicts outlet moisture content (MPRFNN), and the other predicts the discharge rate of the dryer (RPRFNN). One PI controller adjusts the objective of the discharge rate by using MPRFNN, and the other adjusts the given frequency of the discharge motor to control the discharge rate of the grain dryer to reach its objective by using RPRFNN. The experiment is carried out by applying the proposed scheme on the control of a gain dryer with four stages to confirm its effectiveness.

This paper presents a TSK-type recurrent fuzzy neural network (TRFNN) system and hybrid algorithm to control nonlinear uncertain systems. The TRFNN is modified from the RFNN to obtain generalization and fast convergence rate. The consequent part is replaced by linear combination of input variables and the internal variable- fire strength is feedforward to output to increase the network ability. Besides, a hybrid learning algorithm (GA_BPPSO) is proposed to increase the convergence, which combines the genetic algorithm (GA), back-propagation (BP), and particle swarm optimization (PSO). Several simulation results are proposed to show the effectiveness of TRFNN system and GA_BPPSO algorithm.

A GA-based adaptive fuzzy-neural controller for a class of multi-input multi-output nonlinear systems, such as robotic systems, is developed for using observers to estimate time derivatives of the system outputs. The weighting parameters of the fuzzy-neural controller are tuned on-line via a genetic algorithm (GA). For the purpose of on-line tuning the weighting parameters of the fuzzy-neural controller, a Lyapunov-based fitness function of the GA is obtained. Besides, stability of the closed-loop system is proven by using strictly-positive-real (SPR) Lyapunov theory. The proposed overall scheme guarantees that all signals involved are bounded and the outputs of the closed-loop system track the desired output trajectories. Finally, simulation results are provided to demonstrate robustness and applicability of the proposed method.

A new method for active noise control is proposed and experimentally demonstrated. The method is based on Adaptive Neuro-Fuzzy Inference Systems (ANFIS), which is introduced to overcome nonlinearity inherent in active noise control. A new algorithm referred to as Filtered-X ANFIS algorithm suitable for active noise control is proposed. Real-time experiment of Filtered-X ANFIS is performed using floating point Texas Instruments C6701 DSP. In contrast to previous work on ANC using computational intelligence approaches which concentrate on single channel and off-line adaptation, this research addresses multichannel and employs online adaptation, which is feasible due to the computing power of the DSP.

Environment model and communication time delays of a teleoperation system are variant usually, which will induce bad performance, even instability of the system. In this paper, neural network based multiple model adaptive predictive control method is proposed to solve this problem. The whole control system is composed of predictive controller and decision controller. First of all, neural network model set of any possible environment is built up, and time forward state observer based predictive controllers are designed for all models. In succession, decision controller is designed to adaptive switch among all predictive controllers according to performance target. This method can ensure stability and performance of the system. Finally, simulation results show effectiveness of the proposed method.

This paper addresses the problem of wing motion control of flapping wing Micro Air Vehicles (MAVs). Inspired by hummingbird’s wing structure as well as the construction of its skeletal and muscular components, a dynamic model for flapping wing is developed. As the model is highly nonlinear and coupled with unmeasurable disturbances and uncertainties, traditional strategies are not applicable for flapping wing motion control. A new approach called neural-memory based control is proposed in this work. It is shown that this method is able to learn from past control experience and current/past system behavior to improve its performance during system operation. Furthermore, much less information about the system dynamics is needed in construction such a control scheme as compared with traditional NN based methods. Both theoretical analysis and computer simulation verify its effectiveness.

In this paper, a robust neural networks adaptive control scheme is proposed for the stabilization of uncertain linear systems with time-varying delay and bounded perturbations. The uncertainty is assumed to be unknown continuous function without norm-bounded restriction. The perturbation is sector-bounded. Combined with liner matrix inequality method, neural networks and adaptive control, the control scheme ensures the stability of the close-loop system for any admissible uncertainty.

Multiple controllers based on multiple radial based function neural network(RBFNN) models are used to control a nonlinear system to trace a set-point. Considering the nonlinearity of the system, when the set-point value is time variant, a controller based on a fixed structure RBFNN can not give a good control performance. A switching controller which switches among different controller based on different RBFNN is used to adapt the varing set-point value and improve the output reponse and control performance of the nonlinear system.

In this paper, a novel adaptive tracking control strategy is established for general non-Gaussian stochastic systems based on two-step neural network models. The objective is to control the conditional PDF of the system output to follow a given target function by using dynamic neural network models. B-spline neural networks are used to model the dynamic output probability density functions (PDFs), then the concerned problem is transferred into the tracking of given weights corresponding to the desired PDF. The dynamic neural networks with undetermined parameters are employed to identify the nonlinear relationships between the control input and the weights. To achieve control objective, an adaptive state feedback controller is given to estimate the unknown parameters and control the nonlinear dynamics.

An scheme of composite sliding control is proposed for a class of uncertainty nonlinear system, which is based on fuzzy neural networks (FNN) and simple neural networks (SNN). The SNN is uniquely determined by the design of the global integral sliding mode surface, the output of which replaces the corrective control, and FNN is applied to mimic the equivalent control. In this scheme, the bounds of the uncertainties and the extern disturbance are not required to be known in advance, and the stability of systems is analyzed based on Lyapunov function. Simulation results are given to demonstrate the effectiveness of this scheme.

In this paper, a robust adaptive control scheme is proposed for a class of uncertain MIMO time delay systems based on backstepping method with Radical basis function(RBF) neural network. The system uncertainty is approximated by RBF neural networks, and a parameter update law is presented for approximating the system uncertainty. In each step, the control scheme is derived in terms of linear matrix inequalities (LMI’s). A robust adaptive controller is designed using backstepping and LMI method based on the output of the RBF neural networks. Finally, an example is given to illustrate the availability of the proposed control scheme.

This paper proposes to use a radial basis function (RBF) neural network in realising an adaptive control law for air/fuel ratio (AFR) regulation of automotive engines. The sliding mode control (SMC) structure is used and a new sliding surface is developed in the paper. The RBF network adaptation and the control law are derived using the Lyapunov method so that the entire system stability and the network convergence are guaranteed. The developed method is evaluated by computer simulation using the well-known mean value engine model (MVEM) and the effectiveness of the method is proved.

In this paper, we first present a series of dynamic TS fuzzy subsystems to approximate a nonlinear singularly perturbed system. Then the reference model with same fuzzy sets is established. To make the states of the closed-loop system follow those of the reference model, a controller including of neuro-fuzzy adaptive and linear feedback term is designed. The linear feedback parameters can be solved by LMI approach. Adaptive term is used to compensate the uncertainty and alleviate the external disturbance. Lyapunov constitute techniques can prove the stability of the closed loop systems. The simulations results illustrate the effectiveness of this approach.

A temperature control solution is proposed in this paper, which uses a self-tuning PID controller based on flexible neural network (FNN). The learning algorithm of FNN can adjust not only the connection weights but also the sigmoid function parameters. This makes FNN characterized with online learning and high learning speed. The FNN has the following advantages when applied to temperature control problems: high learning ability, which considerably reduces the controller training time; the mathematical model of the plant is not required, which eases the design process; high control performance. These advantages are verified by its application to a practical temperature controlled box, which is used in medicinal inspection. The proposed system presents better behavior than that when using traditional back-propagation neural network.

Neural network controller using adaptation algorithm is a new and simple controller, in which a feedback network propagating the error is not required. So it can be applied to hardware easily. Nevertheless, our simulations show that while the order of controlled plant is high, some unstable phenomenon appear and we also find that sometimes the error is far from being satisfactory, although when the order of controlled plant is low. Moreover, the present adaptation algorithm can not solve this problem. In this paper we will give our derivation of adaptation algorithm used in the neural network controller and configuration of an adaptive neural network controller. Then give some simulation figures to illustrate defect for the new controller. Finally we will develop a hybrid neural network to solve the problem and improve the accuracy as well as reduce the cost to the least in the practical application.

This letter extends adaptive neural network control method to a class of stochastic nonlinear output-feedback systems . Differently from the existing results, the nonlinear terms are assumed to be completely unknown and only a neural network is employed to compensate all unknown nonlinear functions. Based on stochastic LaSalle theorem, the resulting closed-loop system is proved to be globally asymptotically stable in probability.

In this paper, adaptive neural network control is proposed for a class of strict-feedback nonlinear time-delay systems. Unknown smooth function vectors and unknown time-delay functions are approximated by two neural networks, respectively, such that the requirement on the unknown time-delay functions is relaxed. In addition, the proposed systematic backstepping design method has been proven to be able to guarantee semiglobally uniformly ultimately bounded of closed loop signals, and the output of the system has been proven to converge to a small neighborhood of the desired trajectory. Finally, simulation result is presented to demonstrate the effectiveness of the approach.

In this paper, a feedforward active noise control (ANC) system using a recurrent fuzzy neural network (RFNN) controller based on simultaneous perturbation stochastic approximation (SPSA) algorithm is considered. Because RFNN can capture the dynamic behavior of a system through the feedback links, only one input node is needed, and the exact lag of the input variables need not be known in advance. The SPSA-based RFNN control algorithm employed in the ANC system is first derived. Following this, computer simulations are carried out to verify that the SPSA-based RFNN control algorithm is effective for a nonlinear ANC system. Simulation results show that the proposed scheme is able to significantly reduce disturbances without the need to model the secondary-path and has better tracking ability under variable secondary-path. This observation implies that the SPSA-based RFNN controller eliminates the need of the modeling of the secondary-path.

This paper presents a neural network based control design to handle the stabilization of a class of multiple input nonlinear systems with time varying uncertain parameters while assuming that the range of each individual uncertain parameter is known. The proposed design approach allows incorporation of complex control performance measures and physical control constraints whereas the traditional adaptive control techniques are generally not applicable. The desired system dynamics are analyzed, and a collection of system dynamics data, that represents the desired system behavior and approximately covers the region of stability interest, is generated and used in the construction of the neural controller based on the proposed neural control design. Furthermore, the theoretical aspects of the proposed neural controller are also studied, which provides insightful justification of the proposed neural control design. The simulation study is conducted on a single-machine infinity-bus (SMIB) system with time varying uncertainties on its parameters. The simulation results indicate that the proposed design approach is effective.

This paper attempts to present a neural inverse control design framework for a class of nonlinear multiple-input multiple-output (MIMO) system with uncertainties. This research effort is motivated by the following considerations: (a) An appropriate reference model that accurately represents the desired system dynamics is usually assumed to exist and to be available, and yet in reality this is not the case often times; (b) In real world applications, there are many cases where controls are constrained within a physically allowable range, which presents another layer of difficulties to directly apply the reference model based inverse control; (c) It is difficult to consider optimal control even for the reference model as in general the analytic solution to the optimal control problem is not available. The simulation study is conducted on a single-machine infinite-bus (SMIB) system to illustrate the proposed design procedure and demonstrates the effectiveness of the proposed control approach.

The position tracking control of a missile electro-hydraulic servo system is studied. Since the dynamics of the system are highly nonlinear and have large extent of model uncertainties, such as big changes in parameters and external disturbance, a design method of sliding mode control (SMC) using recurrent fuzzy neural network (RFNN) is proposed. First a SMC system, which is insensitive to uncertainties including parameter variations and external disturbance, is introduced. Then, to overcome the problems with SMC, such as the assumption of known uncertainty bounds and the chattering phenomena in the control signal, an RFNN is introduced in conventional SMC. An RFNN bound observer is utilized to adjust the uncertainty bounds in real time. Simulation results verify the validity of the proposed approach.

The molten carbonate fuel cell (MCFC) is a complex system, and MCFC modeling and control are very difficult in the present MCFC research and development because MCFC has the complicated characteristics such as nonlinearness, uncertainty and time-change. To aim at the problem, the MCFC mechanism is analyzed, and then MCFC modeling based on feedback neural networks is advanced. At last, as a result of applying the model, a new MCFC control strategy is presented in detail so that it gets rid of the limits of the controlled object, which has the imprecision, uncertainty and time-change, to achieve its tractability and robustness. The computer simulation and the experiment indicate that it is reasonable and effective.

This paper is concerned with the problem of observer-based stabilization for time-delay systems. Both the state delay and input delay under consideration are assumed to be a constant time-delays, but not known exactly. A new design method is proposed for an observer-based controller with adaptation to the time-delays. The designed controller simultaneously contains both the current state and the past information of systems. The design for adaptation law to delay constants is more concise than the existing conclusions. The controller can be derived by solving a set of linear matrix inequalities (LMIs).

A two-stage structure model was developed for the vibration control of an actuator platform and a controller based on a three-layer neural network was applied to realize high performance control for the kickstand disturbance of a block forming machine. This paper presents a survey of the basic theory of the back-propagation(BP) neural network architecture including its architectural design, BP algorithm, the root mean square error (RMSE) and optimal model establishment. The situ-test data of the control system were measured by acceleration transducer and the experimental results indicates that the proposed method was effective.

A class of Internet congestion control algorithms with communication delays is studied. The algorithm is a pieced continuous function that will be switched on the rate of the source. Based on the Lyapunov theorem, the Lyapunov stability of the system is analyzed. By applying Barbalat Lemma, the global asymptotical stability (GAS) of the algorithm is proved, and a more concise criterion is presented.

A class of window-based network congestion control system with communication delays is studied. By analyzing the network system with communication delay, a critical value of the window size to ensure the stability of network is obtained, and a critical value of the delay to ensure the system stability is presented. Enlarging the delay across the critical value, we find that the congestion control system exhibits Hopf bifurcation.

The variable frequency speed-regulating system which consists of an induction motor and a general inverter, and controlled by PLC is widely used in industrial field. However, for the multivariable, nonlinear and strongly coupled induction motor, the control performance is not good enough to meet the needs of speed-regulating. The mathematic model of the variable frequency speed-regulating system in vector control mode is presented and its reversibility has been proved. By constructing a neural network inverse system and combining it with the variable frequency speed-regulating system, a pseudo-linear system is completed, and then a linear close-loop adjustor is designed to get high performance. Using PLC, a neural network inverse system can be realized in actural system. The results of experiments have shown that the performances of variable frequency speed-regulating system can be improved greatly and the practicability of neural network inverse control was testified.

The Loss-Free resistor (LFR) is applied to DC motor speed control system. A compensation Control algorithm based on LFR is proposed. The LFR is realized by means of switching network whose characteristics are nonlinear, thus, a neural network was designed and used. The switching on-off time can be instantaneously calculated by using the Neural-network-based algorithm proposed. The varies of the motor speed is realized by controlling the output of the LFR, the energy loss in the system is reduced compared with using a conventional power amplifier, and the dynamic characteristics of the system are also improved. The validation of the simulation results is proposed as well.

In this paper, a practical adaptive robust nonlinear controller is proposed for motion control of an SISO nonlinear mechanical system, where the distrubances due to ripple force and friction are compensated by the RBF networks. Rigorous analysis of transient performance and ultimate bound is given. Numerical examples are included to verify the theoretical results.

In this note, robust adaptive neural network (NN) control scheme is constructed for a class of unknown nonlinear systems with drift terms. The robust adaptive NN control laws are developed using backstepping technique which does not require the unknown parameters to be linear parametrizable and no regression matrices are needed. All the signals in the resulting closed-loop system are proved to be ultimately uniform bounded, and the system states are guaranteed to converge to zero.

As we know, saturation, deadzone, backlash, and hysteresis are the most common actuator nonlinearities in practical control system applications. Saturation nonlinearity is unavoidable in most actuators. In this paper, we address the Neural Network saturation compensation for a class of switched nonlinear systems with actuator saturation. An actuator saturation compensation switching scheme for switched nonlinear systems with its subsystem in Brunovsky canonical form is presented using Neural Network. The actuator saturation is assumed to be unknown and the saturation compensator is introduced into a feed-forward path. The scheme that leads to switched stability and disturbance rejection is rigorously proved. The tracking performance of switched nonlinear system is guaranteed based on common Lyapunov approach under the designed switching strategy.

Reheat steam circle system is usually used in modern super-high parameters unit of power plant, which has the characteristics of long process channel, large inertia and long time lag, etc. Thus conventional PID control strategy cannot achieve good control performance. Prompted by the feedback regulation mechanism of biology immune response and the virtues of CMAC neural network, a composite control strategy based on CMAC neural network and immune PID controller is presented in this paper, which has the effect of feed-forward control for load changes as the unit load channel signal of reheat steam temperature is transmitted to the CMAC neural network to take charge of load change effects. The input signal of the controlled system are weighted and integrated by the output signals of CMAC neural network and immune PID controller, and then a variable parameter robust controller is constituted to act on the controlled system. Thus, good regulating performance is guaranteed in the initial control stage and also in case of characteristic deviations of the controlled system. Simulation results show that this control strategy is effective, practicable and superior to conventional PID control.

This paper presents the application of a Grey Box Neural Model (GNM) in adaptive-predictive control of the combustion chamber temperature of a pilot-scale vibrating fluidized dryer. The GNM is based upon a phenomenological model of the process and a neural network that estimates uncertain parameters. The GNM was synthesized considering the energy balance and a radial basis function neural network (RBF) trained on-line to estimate heat losses. This predictive model was then incorporated into a predictive control strategy with one step look-ahead. The proposed system shows excellent results with regard to adaptability, predictability and control when subject to setpoint and disturbances changes.

For stochastic systems with non-Gaussian variables, a descriptor nonlinear system model based on linear B-spline approximation is first established. A new tracking strategy based on

H

 ∞ 

state feedback control for the descriptor nonlinear system is proposed, with which the probability density functions (PDFs) tracking control problem of the non-Gaussian stochastic systems can be solved. Necessary and sufficient condition for the existence of

H

 ∞ 

state feedback controller of the problem is presented by linear-matrix-inequality (LMI). Furthermore, simulations on particle distribution control problems are given to demonstrate the efficiency of the proposed approach and encouraging results have been obtained.

The B-spline neural networks are used to model probability density function (PDF) with least square algorithm, the controllers are designed accordingly. Both the modeling and control methods are tested with molecular weight distribution (MWD) through simulation.

This paper presents a model predictive control (MPC) based on a neural network (NN) model for air/fuel ration (AFR) control of automotive engines. The novelty of the paper is that the severe nonlinearity of the engine dynamics are modelled by a NN to a high precision, and adaptation of the NN model can cope with system uncertainty and time varying effects. A single dimensional optimization algorithm is used in the paper to speed up the optimization so that it can be implemented to the engine fast dynamics. Simulations on a widely used mean value engine model (MVEM) demonstrate effectiveness of the developed method.

Dynamic programming (DP) is a useful tool for solving many control problems, but for its complexity in computation, traditional DP control algorithms are not satisfactory in fact. So we must look for a new method which not only has the advantages of DP but also is easier in computation. In this paper, approximate dynamic programming (ADP) based controller system has been used to solve a ship heading angle keeping problem. The ADP controller comprises successive adaptations of two neural networks, namely action network and critic network which approximates the Bellman equations associated with DP. The Simulation results show that the ship keeps the desired heading satisfactorily.

With the discrete traffic model of an oversaturated intersection, the technique of neural dynamic optimization is used to approximate the optimal signal timing strategy which can lead the minimal delay time while considering the whole congestion period. Our approach can provide an approximation of the optimal timing split in each cycle, as well as the most reasonable number of cycles for specific oversaturated traffic inflows. Specifically, for the two-phase case, we are interested to find that the optimal timing strategy is a bang-bang like control instead of a strict bang-bang one as proposed in relative literature. Moreover, our approach is evaluated with a general four-phase case, and its optimal strategy appears also to be a bang-bang like control, which may illuminate the traffic signal timing in practice.

A communication network is a highly complex nonlinear dynamical system. To avoid congestion collapse and keep network utilization high, many congestion control methods have been proposed. In this paper, a new framework, using Adaptive Critic Designs (ACD) based on Approximate Dynamic Programming (ADP) theory, is presented for network congestion control. At the present time, almost all ACD controllers are designed for centralized control system. In the new frame, the whole network is considered as a multiple non-cooperative ACDs control system, wherein, each source controller is governed by an ACD. This frame provides a new approach to solve the congestion control problem of the networks.

This paper discusses a new application of adaptive dynamic programming (ADP). Meanwhile, traffic control as an important factor in social development is a valuable research topic. Considering with advancement of ADP and importance of traffic control, this paper present a new signal control in a single intersection. Simulation results show that the proposed signal control is valid.

An adaptive critic design is used in the nosiheptide fermentation process to solve the intractable optimization problem. The utility function is defined as the increment of biomass concentration at the adjacent intervals. The state variables are chosen as the biomass concentration, the substrate concentration, the dissolved oxygen concentration and the inhibition concentration. The decision variables are chosen as the temperature, the stirring speed, the airflow and the tank pressure. The adaptive critic method determines optimal control laws for a system by successively adapting the critic networks and the action network. The simulation shows at the same initial conditions this technique can make the fermentation shorten 6 hours.

This paper mainly discusses a generic scheme for on-line adaptive critic design for nonlinear system based on neural dynamic programming (NDP), more exactly, an improved action-depended dual heuristic dynamic programming (ADDHP) method. The principal merit of the proposed method is to avoid the model neural network which predicts the state of next time step, and only use current and previous states in the method, as makes the algorithm more suitable for real-time or on-line application for process control. In this paper, convergence proof of the method will also be given to guarantee the control to reach the optimal. At last, simulation result verifies the performance.

We extend a reinforcement learning algorithm, REINFORCE [13] which has previously been used to cluster data [10]. By using base Gaussian learners, we extend the method so that it can perform a variety of unsupervised learning tasks such as principal component analysis, exploratory projection pursuit and canonical correlation analysis.

According to Hebb’s

Cell assembly theory

, the brain has the capability of function localization. On the other hand, it is suggested that the brain has three different learning paradigms: supervised, unsupervised and reinforcement learning. Inspired by the above knowledge of brain, we present a hierarchical learning system consisting of three parts: supervised learning (SL) part, unsupervised learning (UL) part and reinforcement learning (RL) part. The SL part is a main part learning input-output mapping; the UL part realizes the function localization of learning system by controlling firing strength of neurons in SL part based on input patterns; the RL part optimizes system performance by adjusting parameters in UL part. Simulation results confirm the effectiveness of the proposed hierarchical learning system.

This paper proposes novel hierarchical self-organizing associative memory architecture for machine learning. This memory architecture is characterized with sparse and local interconnections, self-organizing processing elements (PE), and probabilistic synaptic transmission. Each PE in the network dynamically estimates its output value from the observed input data distribution and remembers the statistical correlations between its inputs. Both feed forward and feedback signal propagation is used to transfer signals and make associations. Feed forward processing is used to discover relationships in the input patterns, while feedback processing is used to make associations and predict missing signal values. Classification and image recovery applications are used to demonstrate the effectiveness of the proposed memory for both hetero-associative and auto-associative learning.

A novel machine learning paradigm, i.e. enclosing machine learning based on regular geometric shapes was proposed. It adopted regular minimum volume enclosing and bounding geometric shapes (sphere, ellipsoid, and box) or their unions and so on to obtain one class description model and thus imitate the human “Cognizing” process. A point detection and assignment algorithm based on the one class description model was presented to imitate the human “Recognizing” process. To illustrate the concept and algorithm, a minimum volume enclosing ellipsoid (MVEE) strategy for enclosing machine learning was investigated in detail. A regularized minimum volume enclosing ellipsoid problem and dual form were presented due to probable existence of zero eigenvalues in regular MVEE problem. To solve the high dimensional one class description problem, the MVEE in kernel defined feature space was presented. A corresponding dual form and kernelized Mahalanobis distance formula was presented. We investigated the performance of the enclosing learning machine via benchmark datasets and compared with support vector machines (SVM).

In this paper we derive a simple reinforcement learning rule based on a more general form of REINFORCE formulation. We test our new rule on both classification and reinforcement problems. The results have shown that although this simple learning rule has a high probability of being stuck in local optimum for the case of classification tasks, it is able to solve some global reinforcement problems (e.g. the cart-pole balancing problem) directly in the continuous space.

A novel online dynamic value system for machine learning is proposed in this paper. The proposed system has a dual network structure: data processing network (DPN) and information evaluation network (IEN). The DPN is responsible for numerical data processing, including input space transformation and online dynamic data fitting. The IEN evaluates results provided by DPN. A dynamic three-curve fitting (TCF) scheme provides statistical bounds to the curve fitting according to data distribution. The system uses a shift register communication channel. Application of the proposed value system to the financial analysis (bank prime loan rate prediction) is used to illustrate the effectiveness of the proposed system.

Manifold learning algorithms have been proven to be capable of discovering some nonlinear structures. However, it is hard for them to extend to test set directly. In this paper, a simple yet effective extension algorithm called PIE is proposed. Unlike LPP, which is linear in nature, our method is nonlinear. Besides, our method will never suffer from the singularity problem while LPP and KLPP will. Experimental results of data visualization and classification validate the effectiveness of our proposed method.

As an important active defense technique for computer networks, intrusion detection has received lots of attention in recent years. However, the performance of current intrusion detection systems (IDSs) is far from being satisfactory due to the increasing number of complex sequential attacks. Aiming at the above problem, in this paper, a novel kernel-based reinforcement learning method for sequential behavior modeling in host-based IDSs is proposed. Based on Markov process modeling of host-based intrusion detection using sequences of system calls, the performance optimization of IDSs is transformed to a sequential prediction problem using evaluative reward signals. By using the kernel-based learning prediction algorithm, i.e., the kernel least-squares temporal-difference (kernel LS-TD) algorithm, which implements LS-TD learning in a kernel-induced feature space, the nonlinear modeling and prediction problem for sequential behaviors in IDSs is efficiently solved. Experiments on system call data from the University of New Mexico illustrate that the proposed kernel-based RL approach can achieve better detection accuracy than previous sequential behavior modeling methods including Hidden Markov Models (HMMs) and linear TD algorithms.

In electric power system, long term peak load forecasting plays an important role in terms of policy planning and budget allocation. The planning of power system expansion project starts with the forecasting of anticipated load requirement. Accurate forecasting method can be helpful in developing power supply strategy and development plan, especially for developing countries where the demand is increased with dynamic and high growth rate. This paper proposes a peak load forecasting model using relevance vector machine (RVM), which is based on a probabilistic Bayesian learning framework with an appropriate prior that results in a sparse representation. The most compelling feature of the RVM is, while capable of generalization performance comparable to an equivalent support vector machine (SVM), that it typically utilizes dramatically fewer kernel functions. The proposed method has been tested on a practical power system, and the result indicates the effectiveness of such forecasting model.

The forecasting models for stock market index using computational intelligence such as Artificial Neural networks(ANNs) and Genetic programming(GP), especially hybrid Immune Programming (IP) Algorithm and Gene Expression Programming(GEP) have achieved favorable results. However, these studies, have assumed a static environment. This study investigates the development of a new dynamic decision forecasting model. Application results prove the higher precision and generalization capacity of the predicting model obtained by the new method than static models.

All the factors that influence the rolling force are analyzed, and the neural network model which uses the back propagation (BP) learning algorithm for the calculation of rolling force is created. The initial network’s weights corresponding to the input material grades are taught by the traditional theoretical model, and saved in the database. In order to increase the prediction accuracy of rolling force, we use the measured rolling force data to teach the neural network after several coils of the same input material are rolled down.

Normal fuzzy CMAC neural network performs well because of its fast learning speed and local generalization capability for approximating nonlinear functions. However, it requires huge memory and the dimension increases exponentially with the number of inputs. In this paper, we use recurrent technique to overcome these problems and propose a new CMAC neural network, named recurrent fuzzy CMAC (RFCMAC). Since the structure of RFCMAC is more complex, normal training methods are difficult to be applied. A new simple algorithm with a time-varying learning rate is proposed to assure the learning algorithm is stable.

The identification of nonlinear dynamic systems using fuzzy neural networks is studied. A fast recursive algorithm (FRA) is proposed to select both the fuzzy regressor terms and associated parameters. In comparison with the popular orthogonal least squares (OLS) method, FRA can achieve the fuzzy neural modelling with high accuracy and less computational effort.

This paper focuses on seeking an appropriate number of rules for a T-S inference system. A growing and pruning strategy in neural network is employed, which relates one fuzzy rule’s contribution to the modeling accuracy by a statistic criterion, such that fuzzy rules is added/removed, whereas all the parameters can learn using EKF, both absolutely on-line and with small computation. A simulation for nonlinear system identification illustrates the good performance.

Electromagnetic (EM)–based neural network (NN) approaches have recently gained recognition as unconventional and useful methods for radio frequency (RF) components modeling. In this paper, several improvement techniques including a new data preprocessing technique and an improved training algorithm are presented. Comprehensive cases are compared in this paper. The experimental results indicate that with these techniques, the modified model has better performance.

This paper proposes a novel high-order associative memory system (AMS) based on the Newton’s forward interpolation (NFI), The Interpolation Polynomials and training algorithms for the new AMS scheme are derived. The proposed novel AMS is capable of implementing error-free approximation to complex nonlinear functions of arbitrary order. A method Based on NFI-AMS is designed to model and predict network traffic dynamics, which is capable of modeling the complex nonlinear behavior of a traffic time series and capturing the properties of network traffic. The simulation results showed that the proposed scheme is feasible and efficient. Furthermore, the NFI-AMS based traffic prediction can be used in more fields for network design, management and control.

This paper presents a hybrid modeling method which incorporates knowledge-based components elicited from human expertise into underlying data-driven neural-fuzzy network models. Two different methods in which both measured data and a priori knowledge are incorporated into the model building process are discussed. Based on the combination of fuzzy logic and neural networks, a simple and effective knowledge-based neural-fuzzy network model has been developed and applied to the impact toughness prediction of alloy steels. Simulation results show that the model performance can be improved by incorporating expert knowledge into existing neural-fuzzy models.

In the limestone/gypsum wet flue gas desulphurization (FGD) technology, the change of slurry pH value in absorber is a nonlinear and time-variation process with a large number of uncertainties, so it’s difficult to acquire satisfying mathematical model. In this paper, a novel multiple improved PIDNN ensemble model is proposed to establish the model of slurry pH value. In this model, the concepts of variable integral and partial differential are introduced in the design of hidden-layer of PIDNN, and the concept of output feedback is utilized to improve the ability of PIDNN for dynamic modeling, then multiple improved PIDNN are dynamic combined to get the system output. The results of simulation with field data of wet FGD indicate the validity of this modeling approach.

In this paper, we propose an approach for acoustic modeling using Hidden Markov Models (HMMs) in the Mercer kernel feature space. Acoustic modeling of subword units of speech involves classification of varying length sequences of speech parametric vectors belonging to confusable classes. Nonlinear transformation of the space of parametric vectors into a higher dimensional space using Mercer kernels is expected to lead to better separability of confusable classes. We study the performance of continuous density HMMs trained using the varying length sequences of feature vectors obtained from the kernel based transformation of parametric vectors. Effectiveness of the proposed approach to acoustic modeling is demonstrated for recognition of spoken letters in E-set of English alphabet, and for recognition of consonant-vowel type subword units in continuous speech of three Indian languages.

To decrease the influence of fuzzy and uncertain factors on the maintenance decision process of diesel engine, a fuzzy-neural-network-based maintenance decision model for diesel engine is presented in this paper. It can make the maintenance of diesel engine follow the prevention policy and take the technology and economy into account at the same time. In the presented model, the fuzzy logic and neural network is integrated based on the state detection technology of diesel engine. The maintenance decision process of diesel engine is analyzed in detail firstly. Then, the fuzzy neural network model of maintenance decision is established, including an entire network and two module sub-networks, where the improved T-S model is used to simply the structure of neural networks. Finally, an example is given to verify the effective feasibility of the proposed method. By training the network, the deterioration degree of the diesel engine and its parts can be obtained to make the right maintenance decision.

In vehicular traffic modelling, the effect of link capacity on travel times is generally specified through a delay function. In this paper, the Radial Basis Function Neural Network (RBFNN) method, integrated into a dynamic network loading process, is utilized to model delays at a highway node. The results of the model structure have then been compared to evaluate the relative performance of the integrated neural network method.

This paper presents a novel algorithm for point correspondences using spectral graph analysis. Firstly, the correspondence probabilities are computed by using the modes of proximity matrix and the method of doubly stochastic matrix. Secondly, the TPS deformation model is introduced into the field of spectral correspondence to estimate the transformation parameters between two matched point-sets. The accuracy of correspondences is improved by bringing one point-set closer to the other in each iteration with transformation parameters estimated from the current correspondences. Experiments on both real-world and synthetic data show that our method possesses comparatively high accuracy.

In this paper, a time-delay structure is included in the neural network architecture to emulate the memory effects of wideband wireless transmitters. A simplified analysis approach is proposed to illustrate that the Real-Valued Time-Delay Neural Network (RVTDNN) is one of the most promising neural networks for modeling a complex dynamic nonlinear system. Then the RVTDNN is utilized to build the complex signal dynamic behavioral model of a wideband transmitter. Finally, a behavioral model with three-layer RVTDNN is employed in an experimental system to demonstrate the effectiveness of RVTDNNs in mimicking the dynamic behaviors of a wideband wireless transmitter.

This paper presents an improved neural network model interpretating sonar readings to build occupancy grids of mobile robot. The proposed model interprets sensor readings in the context of their space neighbors and relevant successive history readings simultaneously. Consequently the presented method can greatly weaken the effects by multiple reflections or specular reflection. The output of the neural network is the probability vector of three possible status(empty, occupancy, uncertainty) for the cell. As for sensor readings integration, three probabilities of cell’s status are updated by the Bayesian update formula respectively, and the final status of cell is defined by Max-Min principle.Experiments performed in lab environment has shown occupancy map built by proposed approach is more consistent, accurate and robust than traditional method while it still could be conducted in real time.

In this study, a prediction model of plasma enhanced chemical deposition (PECVD) data was constructed by using an adaptive network-based fuzzy inference system (ANFIS). The PECVD process was characterized by means of a Box Wilson statistical experiment. The film characteristics modeled are deposition rate and stored charge. The prediction performance of ANFIS models was evaluated as a function of training factors, including the step-size, type of membership functions, and normalization factor of inputs-output pairs. The effects of each training factor were sequentially optimized. The root mean square errors of optimized deposition rate and charge models were 11.94 Å/min and 1.37 ×10

12

/cm

2

, respectively. Compared to statistical regression models, ANFIS models yielded an improvement of more than 20%. This indicates that ANFIS can effectively capture nonlinear plasma dynamics.

Modeling for the ball mill grinding process is still an imperative but difficult problem for the optimal control of mineral processing industry. Due to the integrated complexities of grinding process (strong nonlinearity, unknown mechanisms, multivariable, time varying parameters, etc.), a hybrid intelligent dynamic model is presented in this paper, which includes a phenomenological ball mill grinding model with a neurofuzzy network to describe the selection function of different operating conditions, a populace balance based sump model, a phenomenological hydrocyclone model with some reasoning rules for its parameters correction and a radius basis function network (RBFN) for fine particle size error compensation. With the production data from a ball mill grinding circuit of an iron concentration plant, the experiments and simulation results show the proposed hybrid intelligent modeling approach effective.

This paper firstly provides a short introduction to least square support vector machine (LS-SVM), then provides sequential minimal optimization (SMO) based on Pruning Algorithms for LS-SVM, and uses LS-SVM to model nonlinear systems. Simulation experiments are performed and indicated that the proposed method provides satisfactory performance with excellent accuracy and generalization property and achieves superior performance to the conventional method based on common LS-SVM and neural networks.

The paper presents a pattern-oriented agent-based model to simulate the dynamics of a stock market. The model generates satisfactory market macro-level trend and volatility while the agents obey simple rules but follow the behaviors of the neighbors closely. Both the market and the agents are made to evolve in an environment where Darwin’s natural selection rules apply.

Estimating the non-flat function which comprises both the steep variations and the smooth variations is a hard problem. The existing kernel methods with a single common variance for all the regressors can not achieve satisfying results. In this paper, a novel multi-scale model is constructed to tackle the problem by orthogonal least squares regression (OLSR) with wavelet kernel. The scheme tunes the dilation and translation of each wavelet kernel regressor by incrementally minimizing the training mean square error using a guided random search algorithm. In order to prevent the possible over-fitting, a practical method to select termination threshold is used. The experimental results show that, for non-flat function estimation problem, OLSR outperforms traditional methods in terms of precision and sparseness. And OLSR with wavelet kernel has a faster convergence rate as compared to that with conventional Gaussian kernel.

Sensorlesstension control of multi-motor synchronous system with closed tension loop is required in many fields. How to identify the knowledge of instantaneous magnitude of tension is key. In this paper the tension identification is managed on the base of stator currents and its previous values with neural network. According to the fundamental state equations of multi-motor system for tension control, the novel method of tension identification using neural network is presented .A multi-layer feed-forward neural network (MFNN) is trained by Back Propagation Levenberger-Marquardt’s method. Simulation and experiment results show that the system with tension identification via a neural network has better performance, and it can be used in many application fields.

Operon represents a basic organizational unit in microbial genomes. Operon prediction is an important step to study genic transcriptional and regulatory mechanism in microbial genomes. This paper predicted operons in the Escherichia coli K12 genome using neural network based on four types of genomic log-likelihood data. First this method estimated the log-likelihood values for intergenic distances, COG gene functions, conserved gene pairs and phylogenetic profiles, and then used these information by a generalized regression neural network to discriminate pairs of genes within operons (WO pairs) or transcription unit borders (TUB pairs). We test the method on E. coli K12 and find that it can obtain average sensitivity, specificity and accuracy at 85.9%, 89.2% and 87.9% respectively, which indicates that the proposed method has a powerful capability for operon prediction.

Due to its structural simplicity and good properties, radial basis function (RBF) neural network has increasingly been used in many areas for the solution of difficult real-world problems, especially the nonlinear system dynamic modeling. However, the major problem toward using RBF network is the appropriate selection of radial basis function parameters. The basis function parameters are in general the centers and the widths. Our attention in this paper is focused on the configuring the optimal set of parameters to make the networks small and efficient based on rough set theory (RST), which is a valid mathematical tool to perform data reduction. RST is first applied to extract the underlying rules from data. The condition components of the rules are then mapped into network centers. For improve performance, the width parameter of each hidden neuron is initialized individual. The valid of this algorithm is illustrated by an example on the modeling of a ship synchronous generator.

To acquire a satisfied accelerometer dynamic compensation effect, the accelerometer model should be with high precision using the traditional method of zero-pole assignment (ZPS). But in the accelerometer output, the noises, the drift error and the disturbances of the system which make the low precision of the built accelerometer model, can not be avoidable. In this paper, an accelerometer dynamic compensation method based on CMAC neural network is proposed. In this method, a dynamic compensation model can be set up according to the measurement data of dynamic response of the accelerometer without knowing its dynamic model. The dynamic compensation model parameters are trained by CMAC neural network. To a kind of micro-silicon piezoresistance accelerometer, the simulation results show that the proposed new dynamic compensation method has the advantages of fast training process, high precision and easy realization of the dynamic compensation device.

A novel multi-input, multi-output generalized radial basis function (RBF) neural networks for nonlinear system modelling is presented in the paper, which uses extend Kalman filter to sequentially update both the output weights and the centers of the network. Simultaneously, such RBF models employ radial basis functions whose form is determined by admissible exponential generator functions. To test the validity of the proposed method, this paper demonstrates that generalized RBF neural networks with the extended Kalman filter can be used effectively for the identification and modelling of nonlinear dynamical systems. Simulation results reveal that the new generalized RBF networks guarantee faster learning and very satisfactory function approximation capability in modeling nonlinear dynamic systems.

For the practical use of brain-computer interface systems, one of the most significant problems is the generalizing ability of the classifiers, since the states of both people and instruments are altering as time goes on. In this paper, a novel chaotic neural network termed KIII model, is introduced to classify single-trial ECoG during motor imagery, acquired in two different sessions. Then, by comparing with other three traditional classifiers, KIII model shows a greater ability to generalize, which demonstrates that KIII model is an effective tool for brain-computer interfaces systems.

This paper aims to plan the obstacle-avoiding path for mobile robots based on the Artificial Immune Algorithm (AIA) developed from the immune principle; AIA has a strong parallel processing, learning and memorizing ability. This study will design and control a mobile robot within a limited special scale. Through a research method based on the AIA, this study will find out the optimum obstacle-avoiding path. The main purpose of this study is to make it possible for the mobile robot to reach the target object safely and successfully fulfill its task through optimal path and with minimal rotation angle and best learning efficiency. In the end, through the research method proposed and the experimental results, it will become obvious that the application of the AIA after improvement in the obstacle-avoiding path planning for mobile robots is really effective.

Path planning is an important task in mobile robot control. When the robot must move rapidly from any arbitrary start positions to any target positions in environment, a proper path must avoid both static obstacles and moving obstacles of arbitrary shape. In this paper, an obstacle avoidance path planning approach for mobile robots is proposed by using Ant-Q algorithm. Ant-Q is an algorithm in the family of ant colony based methods that are distributed algorithms for combinatorial optimization problems based on the metaphor of ant colonies. In the simulation, we experimentally investigate the sensitivity of the Ant-Q algorithm to its three methods of delayed reinforcement updating and we compare it with the results obtained by other heuristic approaches based on genetic algorithm or traditional ant colony system. At last, we will show very good results obtained by applying Ant-Q to bigger problem: Ant-Q find very good path at higher convergence rate.

This paper proposes an algorithm to detect the obstacles in outdoor unstructured environment with monocular vision. It makes use of motion cues in the video streams. Firstly, optical flow at feature points is calculated. Then rotation of the camera and FOE(focal of expansion) are evaluated separately. A non-linear optimization method is adopted to refine the rotation and FOE. Finally, we get inverse TTC(time to contact) with rotation and FOE and detect the obstacles in the scene. The algorithm doesn’t need any assumption that the ground is flat or partially flat as the conventional methods. So it is suitable for outdoor unstructured environment. Qualitative and quantitative experiment results show that our algorithm works well on different kinds of terrains.

This paper proposes a novel attention selection system with competition neural network supervised by visual memory. As compared with others, this system can not only attend some salient regions randomly according to sensory information but also mainly focus on some learned objects by the visual memory. So it can be applied in robot self-localization or object tracking. The weights of neural networks can be adapted in real time to environment change.

This paper presents a novel omni-directional mobile robot with obstacle avoidance and self-location function. Since a special transmission mechanism is designed, we introduce the mechanism design of the mobile robot and analyze its kinematics. A fuzzy-neural control algorithm is presented to realize the obstacle avoidance of the mobile robot. The obstacle avoidance system based on FKCN and the control algorithm for obstacle avoidance are described. A self-localization technique that uses the ceiling light for the landmark is proposed. Using this self-location function, the mobile robot could locate itself in a world coordinate system.

This paper investigates the use of SOM to process the signal of a 2D laser scanner encountered in feature extraction (corner) and mobile robot self-localization in indoor environments. It presents the method of combining SOM with occupancy grid matching to improve the self-localization performance at the lower computational cost. Experimental results demonstrate that this method can reliably extract the feature of corner point and can effectively improve the self-localization performance of mobile robot.

A robust adaptive control based on generalized dynamic fuzzy neural network (GD-FNN) is presented for robot manipulators. Fuzzy control rules can be generated or deleted automatically according to their significance to the control system, and no predefined fuzzy rules are required. Being use of radial basis function neural network (RBFNN) the learning speed is very fast. The asymptotic stability of the control system is established using Lyapunov theorem. Simulations are given for a two-link robot in the end of paper, and validated the control arithmetic.

Due to the time-consuming calculation for the forward kinematics of a 3-PRS (prismatic-revolute-spherical) parallel manipulator, neither the kinematic nor dynamic control algorithm can be implemented on real time. To deal with such problem, the forward kinematics is solved by means of artificial neural network (NN) approach in this paper. Based on the trained NN, the kinematic control of the manipulator is carried out by resorting to an ordinary control algorithm. Simulation results illustrate that the NN can approximate the forward kinematics perfectly, which leads to ideal control results of the parallel manipulator.

In a sinusoid like curve configuration, the snake-like manipulator (also called snake arm) has a wide range of potential applications for its redundancy to overcome conventional industrial robot’s limitation when carrying out a complex task. It can perform many kinds of locomotion like the nature snake or the animal’s tentacle to avoid obstacles, follow designated trajectories, and grasp objects. Effectively control of the snake-like manipulator is difficult for its redundancy. In this study, we propose an approach based on BP neural network to kinematic control the hyper-redundant snake-like manipulator. This approach, inspired by the Serpenoid curve and the concertina motion principle of the nature snake, is completely capable of solving the control problem of a planar snake-like manipulator with any number of links following any desired direction and trajectory. With shape transformation and base rotation, the manipulator’s configuration changes accordingly and moves actively to perform the designated tasks. By using BP neural networks in modeling the inverse kinematics, this approach has such superiorities as few control parameters and high precision. Simulations have demonstrated that this control technique for the snake-like manipulator is available and effective.

Multi-Constrained Shortest Path (MCSP) selection is a fundamental problem in communication networks. Since the MCSP problem is NP-hard, there have been many efforts to develop efficient approximation algorithms and heuristics. In this paper, a new algorithm is proposed based on vectorial Autowave-Competed Neural Network which has the characteristics of parallelism and simplicity. A nonlinear cost function is defined to measure the autowaves (

i.e.

, paths). The

M

-paths limited scheme, which allows no more than

M

autowaves can survive each time in each neuron, is adopted to reduce the computational and space complexity. And the proportional selection scheme is also adopted so that the discarded autowaves can revive with certain probability with respect to their cost functions. Those treatments ensure in theory that the proposed algorithm can find an approximate optimal path subject to multiple constraints with arbitrary accuracy in polynomial-time. Comparing experiment results showed the efficiency of the proposed algorithm.

This paper presents a neuro intelligent virtual leader based approach for close formation of a group of mobile vehicles. Neural Network-based trajectory planning is incorporated into the leading vehicle so that an optimal reference path is generated automatically by the virtual leader, which guides the whole team vehicles to the area of interest as precisely as possible. The steering control scheme is derived based on the structural properties of the vehicle dynamics. Simulation on multiple vehicles formation is conducted as a verification of the effectiveness of the proposed method.

This paper puts forward a multi-stages competitive neural networks approach for motion trajectory pattern analysis and learning. In this method, the rival penalized competitive learning method, which could well overcome the competitive networks’ problems of the selection of output neurons number and weight initialization, is used to discover the distribution of the flow vectors according to the trajectories’ time orders. The experiments on different sites with CCD and infrared cameras demonstrate that our method is valid for motion trajectory pattern learning and can be used for anomaly detection in outdoor scenes.

A robust tracking controller with bound estimation based on neural network is proposed to deal with the unknown factors of nonholonomic mobile robot, such as model uncertainties and external disturbances. The neural network is to approximate the uncertainties terms and the interconnection weights of the neural network can be tuned online. And the robust controller is designed to compensate for the approximation error. Moreover, an adaptive estimation algorithm is employed to estimate the bound of the approximation error. The stability of the proposed controller is proven by Lyapunov function. The proposed neural network-based robust tracking controller can overcome the uncertainties and the disturbances. The simulation results demonstrate that the proposed method has good robustness.

NNPC has been used widely to control nonlinear systems. However traditional gradient decent algorithm (GDA) needs a large computational cost, so that NNPC is not acceptable for systems with rapid dynamics. To apply NNPC in fast control of mobile robots, the paper proposes an improved optimization technique, particle swarm optimization with controllable random exploration velocity (PSO-CREV), to replace of GDA in NNPC. Therefore for one cycle of control, PSO-CREV needs less iterations than GDA, and less population size than conventional PSO. Hence the computational cost of NNPC is reduced by using PSO-CREV, so that NNPC using PSO-CREV is more feasible for the control of rapid processes. As an example, a test of trajectory tracking using mobile robots is chosen to compare performance of PSO-CREV with other algorithms to show its advantages, especially on the aspect of computational time.

This paper presents a novel fuzzy-neural control algorithm to realize obstacle avoidance of a mobile robot. A heuristic fuzzy-neural network is developed based on heuristic fuzzy rules and the Kohonen clustering network. By applying the off-line and unsupervised training method to this network, the pattern mapping relation between ultrasonic sensory input and velocity command is established. This paper describes mechanical design of the mobile robot, the arrangement of ultrasonic sensors, the obstacle avoidance system based on FKCN, classification of obstacle, the control algorithm for obstacle avoidance and training data library. In order to verify the effectiveness of this algorithm, we give the results of simulation in a computer virtual environment.

Regression analysis between features of high-dimension is receiving attention in environmental learning of mobile robot. In this paper, we propose a novel framework, namely General regression neural network (GRNN), for approximating the functional relationship between high-dimensional map features and robot’s states. We firstly adopt PCA to preprocess images taken from omnidirenctional vision. The method extracts map features optimally and reduces the correlated features while keeping the minimum reconstruction error. Then, the robot states and corresponding features of the training panoramic snapshots are used to train the given neural network. This enables robot to memorize the environmental features as well as to predict available scene given its location information. Experimental results are shown finally.

The paper proposed a method for a quadruped robot control system based Central Pattern Generator (CPG) and fuzzy neural networks (FNN). The common approach for the control of a quadruped robot includes two methods mainly. One is the CPG that is based the bionics, the other is the dynamic control that is based the model of quadruped robot. The control result of CPG is decided by the gait data of the quadruped and the parameters of the CPG are choosing manually. Modeling a quadruped robot is difficult because it is a high nonlinear system. This paper presents a much simpler method for the control of a quadruped robot. A simple CPG is adopted for a timing oscillator; it generates the motion periodic pattern of legs. The FNN is used to control the joint motion in order to get a desired stable trajectory motion.

In cooperative multiple robot systems (CMRS), dynamic knowledge representation and inference is the key in scheduling robots to fulfill the cooperation requirements

.

The first goal of this work is to use rough set based rules generation method to extract dynamic knowledge of our CMRS. Kang’s rough set based rules generation method is used to get fuzzy dynamic knowledge from practical decision data. The second goal of this work is to use Fuzzy Neural Petri nets (FNPN) to represent and infer the dynamic knowledge on the base of dynamic knowledge extraction with self-learning ability. In particular, we investigate a new way to extract, represent and infer dynamic knowledge with self-learning ability in CMRS. Finally, the effectiveness of the dynamic knowledge extraction, representation and inference procedure are demonstrated.

This paper presents a method of force/position control by using the backstepping and passivity strict-feedback neural networks technique; passivity monitor can evaluate stability of a system based on the concept of passivity. The parameters estimation for the design is made by the neural networks technology, using the decouple method and matrix transforming technology, decomposing the robot system as the position subsystem and the force subsystem, then the control law of these subsystems are designed respectively. The results obtained are satisfactory by using hybrid force and position control, the error is negligible and the global stability of the system can also be obtained.

This paper investigates the problem of global asymptoticstability for a class of uncertain neural networks with time-varying and distributed delays. The uncertainties we considered in this paper are norm-bounded, and possibly time-varying. By Lyapunov-Krasovskii functional approach and S-procedure, a new stability criteria for the asymptotic stability of the system is derived in terms of linear matrix inequalities (LMIs). Two simulation examples are given to demonstrate the effectiveness of the developed techniques.

This paper discusses a mathematical model of network, which is more general than the cellular neural networks(CNNs). In this study, we discuss some dynamical properties of this type of network, such as the distribution of equilibrium points and the influence of external input on stability. Moreover, we give some criterions, which ensure the complete stability of this network.

In this paper, we extend the Cohen–Grossberg neural networks from classical to fuzzy sets, and propose the fuzzy Cohen–Grossberg neural networks (FCGNN). The global exponential stability of FCGNN with time-varying delays is studied. Without assuming the boundedness and differentiability of the activation functions, based on the properties of M-matrix, by constructing vector Liapunov functions and applying differential inequalities, the sufficient conditions ensuring existence, uniqueness, and global exponential stability of the equilibrium point of fuzzy Cohen–Grossberg neural networks with variable delays are obtained.

Locally and globally asymptotical stability on equilibria of delayed neural networks with saturation activation functions are studied by the Razumikhin-type theorems, which are the main approaches to study the stability of functional differential equations, and some new stability conditions are obtained, which are constructed by the networks’ parameters. In the case of local stability conditions, the attracted fields of equilibria are also estimated. All results obtained in this paper need only to compute the eigenvalues of some matrices or to verify some inequalities to be holden.

This paper considers the problem of finite-time boundedness (FTB) of the general delayed neural networks with norm-bounded parametric uncertainties. The concept of FTB for time delay system is extended first. Then, based on the Lyapunov function and linear matrix inequality (LMI) technique, some delay-dependent criteria are derived to guarantee FTB. The conditions can be reduced to a feasibility problem involving linear matric inequalities (LMIs). Finally, two examples are given to demonstrate the validity of the proposed methodology.

In this paper, we study the global asymptotic stability properties of cellular neural networks with variable coefficients and time varying delays. We present sufficient conditions for the global asymptotic stability of the neural networks . The proposed conditions, which are applicable to all continuous nonmonotonic neuron activation functions and do not require the interconnection matrices to be symmetric, establish the relationships between network parameters of the neural systems and the delay parameters. Some examples show that our results are new and improve the previous results derived in the literature.

Discrete-time Cohen-Grossberg neural networks(CGNNs) are studied in this paper. Several sufficient conditions are obtained to ensure the global exponential stability of the discrete-time systems of CGNNs with delays based on Lyapunov methods. The obtained results have not assume the symmetry of the connection matrix, and monotonicity, boundness of the activation functions.

Continuous attractor is a promising model for describing the encoding of continuous stimuli in neural systems. In a continuous attractor, the stationary states of the neural system form a continuous parameter space, on which the system is neutrally stable. This property enables the neutral system to track time-varying stimulus smoothly. In this study we investigate the tracking speed of continuous attractors. In order to analyze the dynamics of a large-size network, which is otherwise extremely complicated, we develop a strategy to reduce its dimensionality by utilizing the fact that a continuous attractor can eliminate the input components perpendicular to the attractor space very quickly. We therefore project the network dynamics onto the tangent of the attractor space, and simplify it to be a one-dimension Ornstein-Uhlenbeck process. With this approximation we elucidate that the reaction time of a continuous attractor increases logarithmically with the size of the stimulus change. This finding may have important implication on the mental rotation behavior.

In this paper, the global asymptotic stability of Hopfield neural networks with time delays is investigated. Some novel sufficient conditions are presented for the global stability of a given delayed Hopfield neural networks by constructing Lyapunov functional and using some well-known inequalities. A linear matrix inequality (LMI) approach is developed to establish sufficient conditions for the given neural networks. An illustrative example is provided to demonstrate the effectiveness of our theoretical results.

In this paper, the periodic solution for a class of Cohen-Grossberg neural networks with variable coefficients is discussed. By using inequality analysis technique and matrix theory, some new sufficient conditions are obtained to ensure the existence, uniqueness, global attractivity and exponential stability of the periodic solution. An example is given to show the effectiveness of the obtained results.

The paper investigates the existence and global stability of periodic solution of non-autonomous neural networks with delay. Then the existence and uniqueness of periodic solutions of the neural networks are discussed in the paper. Moreover, criterion on stability of periodic solutions of the neural networks is obtained by using matrix function inequality, and algorithm for the criterion on the neural networks is provided. Result in the paper generalizes and improves the result in the existing references. In the end, an illustrate example is given to verify our results.

In this paper, a class of generalized nonautonomous cellular neural networks with time-varying delays are studied. By means of Lyapunov functional method, improved Young inequality

a

m

b

n

 ≤ 

ma

t

− 

n

 + 

nb

t

m

(0 ≤ 

m

 ≤ 1,

m

 + 

n

 = 1,

t

 > 0) and the homeomorphism theory, several sufficient conditions are given guaranteeing the existence, uniqueness and global exponential stability of the equilibrium point. The proposed results generalize and improve previous works. An illustrative example is also given to demonstrate the effectiveness of the proposed results.

The global asymptotic stability of discrete-time Cohen–Grossberg neural networks (CGNNs) with or without time delays is studied in this paper. The CGNNs are transformed into discrete-time interval systems, and several sufficient conditions of asymptotic stability for these interval systems are derived by constructing some suitable Lyapunov functionals. The obtained conditions are given in the form of linear matrix inequalities that can be checked numerically and very efficiently by resorting to the MATLAB LMI Control Toolbox.

In this paper, the global asymptotic and exponential stability are investigated for a class of neural networks with distributed time-varying delays. By using appropriate Lyapunov-Krasovskii functional and linear matrix inequality (LMI) technique, two delay-dependent sufficient conditions in LMIs form are obtained to guarantee the global asymptotic and exponential stability of the addressed neural networks. The proposed stability criteria do not require the monotonicity of the activation functions and the differentiability of the distributed time-varying delays, which means that the results generalize and further improve those in the earlier publications. An example is given to show the effectiveness of the obtained condition.

In finite mixture modelling, it is crucial to select the number of components for a data set. We have proposed an entropy regularized likelihood (ERL) learning principle for the finite mixtures to solve this model selection problem under regularization theory. In this paper, we further give an asymptotic analysis of the ERL learning, and find that the global minimization of the ERL function in a simulated annealing way (i.e., the regularization factor is gradually reduced to zero) leads to automatic model selection on the finite mixtures with a good parameter estimation. As compared with the EM algorithm, the ERL learning can go across the local minima of the negative likelihood and keep robust with respect to initialization. The simulation experiments then prove our theoretic analysis.

The existence and global asymptotic stability of a large class of Cohen-Grossberg neural networks is discussed in this paper. Previous papers always assume that the amplification function has positive lower and upper bounds, which excludes a large class of functions. In our paper, it is only needed that the amplification function is positive. Also, the model discussed is general, the method used is direct and the conditions needed are weak.

This paper studies a general class of neural networks with time-varying delays and the neuron activations belong to the set of discontinuous monotone increasing functions. The discontinuities in the activations are an ideal model of the situation where the gain of the neuron amplifiers is very high. Because the delay in combination with high-gain nonlinearities is a particularly harmful source of potential instability, in the paper, conditions which ensure the global convergence of the neural network are derived.

In this paper, we study the Cohen-Grossberg neural networks with mixed time-varying delays. By applying the Lyapunov functional method and combining with the inequality 3

abc

 ≤ 

a

3

 + 

b

3

 + 

c

3

(

a

,

b

,

c

 > 0) technique, a series of new and useful criteria on the existence of equilibrium point and its global asymptotical stability are established. The results obtained in this paper extend and generalize the corresponding results existing in previous literature.

In this paper, the difference between using the inverted neuron output and negative resistor for expressing inhibitory synapse is studied. We analyzed that the total conductance seen at the neuron input is different in these two methods. And this total conductance has been proved to effect on the system stability in this paper. Also, we proposed the method how to stabilize the input space and improve the system’s performance by adjusting the input conductance between neuron input and ground. Pspice is used for circuit level simulation.

In this paper, the global uniform asymptotical stability is studied for neural networks with multiple time-varying delays by constructing appropriate Lyapunov-Krasovskii functional and using the linear matrix inequality (LMI) approach. The restriction on the derivative of the time-varying delay function

τ

ij

(

t

) to be less than unit is removed by using slack matrix method. A numerical example is provided to demonstrate the effectiveness and applicability of the proposed criteria.

In this paper, we investigate dynamical behavior of a general class of competitive or cooperative Lotka-Volterra systems with delays. Positive solutions and global stability of nonnegative equilibrium are discussed. Sufficient condition independent of delays guaranteeing existence of globally stable equilibrium is given. A Simulation verifying theoretical results is given, too.

We present an efficient stream authentication scheme that improves the verification probability of Park-Chung-Cho’s scheme [1] by using Rabin’s Information Dispersal Algorithm. It is shown that under the same communication overhead the verification probability of the proposed scheme is higher than those of SAIDA as well as Park-Chung-Cho’s scheme, and that the execution time of the proposed scheme is smaller than that of SAIDA.

Discrete-time analogues of continuous-time neural networks with continuously distributed delays and periodic inputs are introduced. The discrete-time analogues are considered to be numerical discretizations of the continuous-time networks and we study their dynamical characteristics. By employing Halanay-type inequality, we obtain easily verifiable sufficient conditions ensuring that every solutions of the discrete-time analogue converge exponentially to the unique periodic solutions. It is shown that the discrete-time analogues preserve the periodicity of the continuous-time networks.

This paper proposes a novel artificial neural oscillator consisted of two neurons with excellent control properties. The mutual connections between the neurons are just linear functions and determine the oscillation angular frequency. And each neuron has a nonlinear self-feedback connection to hold up oscillation amplitude. The dynamics of the neural oscillator was modelled with nonlinear coupling functions. And the stability, amplitude, angular frequency of the oscillator are determined independently by three parameters of the functions. Since it has simple structure and favorable control advantages, it can be used in bionic robot’s locomotion control system. The first application is an artificial central pattern generator (CPG) controller for bionic robot’s joint. The second is a bionic neural network for fish-robot’s locomotion control.

Extreme learning machine (ELM) is a novel learning algorithm much faster than the traditional gradient-based learning algorithms for single-hidden-layer feedforward neural networks (SLFNs). Neural network ensemble is a learning paradigm where several neural networks are jointly used to solve a problem. In our work, we investigated the performance of ELMs ensemble on regression problems. A simple ensembling approach Product Index based Excluding ensemble(PIEx) was proposed to ensemble accurate and diverse member networks. The experimental results show that the ensemble can effectively improve the performance compared with the generalization ability of single ELM and PIEx outperforms Bagging and Simple Averaging. The results also show ELM training can generate diverse neural networks even though using the same training set.

Online-sequential extreme learning machine (OS-ELM) shows a good solution to online learning using extreme learning machine approach for single-hidden-layer feedforward network. However, the algorithm tends to be data-dependent, i.e. the bias values need to be adjusted depending on each particular problem. In this paper, we propose an enhancement to OS-ELM, which is referred to as robust OS-ELM (ROS-ELM). ROS-ELM has a systematic method to select the bias that allows the bias to be selected following the input weights. Hence, the proposed algorithm works well for every benchmark dataset. ROS-ELM has all the pros of OS-ELM, i.e. the capable of learning one-by-one, chunk-by-chunk with fixed or varying chunk size. Moreover, the performance of the algorithm is higher than OS-ELM and it produces a better generalization performance with benchmark datasets.

This paper presents an efficient online sequential learning algorithm for extreme learning machine, which can learn data one by one. In this algorithm, the parameters of hidden nodes (the input weights and biases of additive nodes or the centers and impact factors of RBF nodes) are randomly selected and the output weights are analytically determined based on the sequentially arriving data. In the online sequence, the algorithm updates the output-layer weights with a Givens QR decomposition based on the orthogonalized least square algorithm. Simulations on benchmark problems demonstrate that the algorithm produces much better generalization performance than another online sequential extreme learning machine algorithm, or sometimes it has good performance than primitive extreme learning machine algorithm.

Machine learning, a cornerstone of intelligent systems, has typically been studied in the context of specific tasks, including clustering (unsupervised learning), classification (supervised learning), and control (reinforcement learning). This paper presents a learning architecture within which a universal adaptation mechanism unifies a rich set of traditionally distinct learning paradigms, including learning by matching, learning by association, learning by instruction, and learning by reinforcement. In accordance with the notion of embodied intelligence, such a learning theory provides a computational account of how an autonomous agent may acquire the knowledge of its environment in a real-time, incremental, and continuous manner. Through a case study on a minefield navigation domain, we illustrate the efficacy of the proposed model, the learning paradigms encompassed, and the various types of knowledge learned.

Multiple-instance learning (MIL) is a variation on supervised learning, where the task is to learn a concept given positive and negative bags of instances. In this paper a novel algorithm has been introduced for multiple-instance learning. This method was inspired by both diverse density (DD) and its expectation maximization version (EM-DD). It converts MIL problem to a single-instance setting. This improved method has better accuracy and time complexity than DD and EM-DD. We apply it to drug activity prediction and image retrieval. The experiments show it has competitive accuracy values compared with other previous approaches.

Investigated in this paper are the uniform approximation capabilities of sum-of-product (

SOPNN

) and sigma-pi-sigma (

SPSNN

) neural networks. It is proved that the set of functions that are generated by an

SOPNN

with its activation function in

C

(ℝ) is dense in

$C(\mathbb{K})$

for any compact

$\mathbb{K}\in \mathbb{R}^N$

, if and only if the activation function is not a polynomial. It is also shown that if the activation function of an

SPSNN

is in

C

(ℝ), then the functions generated by the

SPSNN

are dense in

$C(\mathbb{K})$

if and only if the activation function is not a constant.

This paper presents a new method of regularization in regression problems using a Besov norm (or semi-norm) acting as a regularization operator. This norm is more general smoothness measure to approximation spaces compared to other norms such as Sobolev and RKHS norms which are usually used in the conventional regularization methods. In our work, we also suggest a new candidate of the regularization parameter, that is, the trade-off between the data fit and the smoothness of the estimation function. Through the simulation for function approximation, we have shown that the suggested regularization method is effective and the estimated values of regularization parameters are close to the optimal values associated with the minimum expected risks.

According to analysis of both neuro-electrophysiological experimental data and the biophysical properties of neurons, in early research paper we proposed a new biophysical model that reflects the property of energy coding in neuronal activity. On the based of the above research work, in this paper the proposed biophysical model can reproduce the membrane potentials and the depolarizing membrane current by means of neuro-electrophysiological experimental data. Combination with our previous research results, the proposed biophysical model is demonstrated again to be more effective compared with known biophysical models of neurons.

In the human mind, high-order knowledge is categorically organized, yet the nature of its internal representation system is not well understood. While it has been traditionally considered that there is a single innate representation system in our mind, recent studies suggest that the representational system is a dynamic, capable of adjusting a representation scheme to meet situational characteristics. In the present paper, we introduce a new cognitive modeling framework accounting for the flexibility in representing high-order category knowledge. Our modeling framework flexibly learns to adjust its internal knowledge representation scheme using a meta-heuristic optimization method. It also accounts for the multi-objective and the multi-notion natures of human learning, both of which are indicated as very important but often overlooked characteristics of human cognition.

We extend the statistical neurodynamics to study transient dynamics of sequence processing neural networks with finite dilution, and the theoretical results are supported by extensive numerical simulations. It is found that the order parameter equations are completely equivalent to those of the Generating Functional Method, which means that crosstalk noise follows normal distribution even in the case of failure in retrieval process. In order to verify the gaussian assumption of crosstalk noise, we numerically obtain the cumulants of crosstalk noise, and third- and fourth-order cumulants are found to be indeed zero even in non-retrieval case.

In this paper, a novel elliptical basis function neural networks model (EBFNN) based on a hybrid learning algorithm (HLA) is proposed. Firstly, a geometry analytic algorithm is applied to construct the hyper-ellipsoid units of hidden layer of the EBFNN, i.e., initial the structure of the EBFNN. Then, the hybrid learning algorithm (HLA) is further applied to adjust the centers and the shape parameters. The experimental results demonstrated the proposed hybrid learning algorithm for the EBFNN model is feasible and efficient, and the EBFNN is not only parsimonious but also has better generalization performance than the RBFNN.

Multiple-Instance Learning is increasingly becoming one of the most promiscuous research areas in machine learning. In this paper, a new algorithm named NRBF-MI is proposed for Multi-Instance Learning based on normalized radial basis function network. This algorithm defined Compact Neighborhood of bags on which a new method is designed for training the network structure of NRBF-MI. The behavior of kernel function radius and its influence is analyzed. Furthermore a new kernel function is also defined for dealing with the labeled bags. Experimental results show that the NRBF-MI is a high efficient algorithm for Multi-Instance Learning.

Compared to normal learning algorithms, for example backpropagation, the optimal bounded ellipsoid (OBE) algorithm has some better properties, such as faster convergence, since it has a similar structure as Kalman filter. OBE has some advantages over Kalman filter training, the noise is not required to be Guassian. In this paper OBE algorithm is applied traing the weights of recurrent neural networks for nonlinear system identification

.

Both hidden layers and output layers can be updated. From a dynamic systems point of view, such training can be useful for all neural network applications requiring real-time updating of the weights. A simple simulation gives the effectiveness of the suggested algorithm.

Radial basis function (RBF) networks of Gaussian activation functions have been widely used in many applications due to its simplicity, robustness, good approximation and generalization ability, etc.. However, the training of such a RBF network is still a rather difficult task in the general case and the main crucial problem is how to select the number and locations of the hidden units appropriately. In this paper, we utilize a new kind of Bayesian Ying-Yang (BYY) automated model selection (AMS) learning algorithm to select the appropriate number and initial locations of the hidden units or Gaussians automatically for an input data set. It is demonstrated well by the experiments that this BYY-AMS training method is quite efficient and considerably outperforms the typical existing training methods on the training of RBF networks for both clustering analysis and nonlinear time series prediction.

We usually identify the categories in image databases using some clustering algorithms based on the visual features extracted from images. Due to the well-known gap between the semantic features (e.g., categories) and the visual features, the results of unsupervised image categorization may be quite disappointing. Of course, it can be improved by adding some extra semantic information. Pairwise constraints between some images are easy to provide, even when we have little prior knowledge about the image categories in a database. A semi-supervised learning algorithm is then proposed for unsupervised image categorization based on Gaussian mixture model through incorporating such semantic information into the entropy-regularized likelihood (ERL) learning, which can automatically detect the number of image categories in the database. The experiments further show that this algorithm can lead to some promising results when applied to image categorization.

This paper attempts to evaluate machine learning based approaches to text categorization including NTC without decomposing it into binary classification problems, and presents another learning scheme of NTC. In previous research on text categorization, state of the art approaches have been evaluated in text categorization, decomposing it into binary classification problems. With such decomposition, it becomes complicated and expensive to implement text categorization systems, using machine learning algorithms. Another learning scheme of NTC mentioned in this paper is unconditional learning where weights of words stored in its learning layer are updated whenever each training example is presented, while its previous learning scheme is mistake driven learning, where weights of words are updated only when a training example is misclassified. This research will find advantages and disadvantages of both learning schemes by comparing them with each other

In this paper we apply harmony data smoothing learning on a weighted kernel density model to obtain a sparse density estimator. We empirically compare this method with the least squares cross-validation (LSCV) method for the classical kernel density estimator. The most remarkable result of our study is that the harmony data smoothing learning method outperforms LSCV method in most cases and the support vectors selected by harmony data smoothing learning method are located in the regions of local highest density of the sample.

Code reuse in software reuse has several limitations such as difficulties of understanding and retrieval of the reuse code written by other developers. To overcome these problems, it should be possible to reuse the analysis/design information than source code itself. In this paper, I present analogical matching techniques for the reuse of object models and design patterns. We have suggested the design patterns as reusable components and the representation techniques to store them. The contents of the paper are as follows. 1) Analogical matching functions to retrieve analogous design patterns from reusable libraries. 2) The representation of reusable components to be stored in the library in order to support the analogical matching.

This paper presents an informative gene set selection approach to tumor diagnosis based on the Distance Sensitive Rival Penalized Competitive Learning (DSRPCL) algorithm and redundancy analysis. Since the DSRPCL algorithm can allocate an appropriate number of clusters for an input dataset automatically, we can utilize it to classify the genes (expressed by the gene expression levels of all the samples) into certain basic clusters. Then, we apply the post-filtering algorithm to each basic gene cluster to get the typical and independent informative genes. In this way we can obtain a compact set of informative genes. To test the effectiveness of the selected informative gene set, we utilize the support vector machine (SVM) to construct a tumor diagnosis system based on the express profiles of its genes. It is shown by the experiments that the proposed method can achieve a higher diagnosis accuracy with a smaller number of informative genes and less computational complexity in comparison with the previous ones.

The problem of fault diagnosis of electrical machine has been an ongoing research in power systems. Many machine learning tools have been applied to this problem using static machine learning structures such as neural network, support vector machine that are unable to accommodate new information as it becomes available into their existing models. This paper presents a new method to bushing fault condition monitoring using fuzzy ARTMAP(FAM). FAM is introduced for bushing condition monitoring because it has the ability to incrementally learn information as it becomes available. An ensemble of classifiers is used to improve the classification accuracy of the systems. The testing results show that FAM ensemble gave an accuracy of 98.5%. Furthermore, the results show that fuzzy ARTMAP can update its knowledge in an incremental fashion without forgetting previously learned information.

A new weighted wavelet neural network is presented, And the approximation capability of such weighted wavelet neural network is also studied based on the traits of Lebesgue partition, the operator theory and the topology structure of the relatively compact set in Hilbert space. The simulation results indicate that the weighted wavelet neural network is a uniformed approximator, which can approximates the nonlinear function in compact set by arbitrary precision.

A method of moving horizon state estimation (MHSE) including a recurrent neural network as the dynamic model is used as an estimator of the filling level of the mill for a semiautogenous ore grinding process. The results are compared to those of a simple neural network acting as an estimator. They show the advantages of the Neural-MHSE, especially concerning robustness under large perturbations of the state variables (index of agreement > 0.9), which would favor its application to industrial scale processes.

Data Mining is an analytic process designed to explore data (usually large amounts of data - typically business or market related) in search of consistent patterns and/or systematic relationships between variables, and then to validate the findings by applying the detected patterns to new subsets of data. One of the most commonly used techniques in data mining, Artificial Neural Networks provide non-linear predictive models that learn through training and resemble biological neural networks in structure. This paper deals with a new adaptive neural network model: a feed-forward higher order neural network with a new activation function called neuron-adaptive activation function. Experiments with function approximation and stock market movement analysis have been conducted to justify the new adaptive neural network model. Experimental results have revealed that the new adaptive neural network model presents several advantages over traditional neuron-fixed feed-forward networks such as much reduced network size, faster learning, and more promising financial analysis.

Coronary heart disease (CHD) is a serious disease causing more and more morbidity and mortality. Combining western medicine and Traditional Chinese Medicine (TCM) to heal CHD becomes especially necessary for medical society today. Since western medicine faces some problems, like high cost and more side effects. TCM can be a complementary alternative to overcome these defects. Identification of what syndrome a CHD patient caught has been a challenging issue for medical society because the core of TCM is syndrome. In this paper, we carry out a large-scale clinical epidemiology to collect data with 1069 cases, each of which must be a CHD instance but may be diagnosed as different syndromes. We take blood stasis syndrome (frequency is 69%) as an example, employ four distinct kinds of data mining algorithms: Bayesian model; Neural Network; Support vector machine and Decision trees to classify the data and compare their performance. The results indicated that neural network is the best identifier with 88.6% accuracy on the holdout samples. The next is support vector machine with 82.5% accuracy, a slight higher than Bayesian model with 82.0% counterpart. The decision tree performs the worst, only 80.4%. We conclude that in identifying syndromes in CHD, neural network can provide a best insight to clinical application.

With the vast amount of mobile user information available today, mining knowledge of mobile users is getting more and more important for a mobile commerce (M-commerce) system. Vector space model (VSM) is one of the most popular methods to achieve the above goal. Unfortunately, it can not identify the latent information in the user feature space, which decreases the quality of personalized services. In this paper, we present a concept-lattice based kernel method for mining the hidden user knowledge. The main idea is to employ concept lattice for constructing item proximity matrix, and then embed it into a kernel function, which transforms the original user feature space into a user concept space, and at last, perform personalized services in the user concept space. The experimental results demonstrate that our method is more encouraging than VSM.

Network anomaly detection has been a hot topic in the past years. However, high false alarm rate, difficulties in obtaining exact clean data for the modeling of normal patterns and the deterioration of detection rate because of “unclean” training set always make it not as good as we expect. Therefore, we propose a novel data mining method for network anomaly detection in this paper. Experimental results on the well-known KDD Cup 1999 dataset demonstrate it can effectively detect anomalies with high true positives, low false positives as well as with high confidence than the state-of-the-art anomaly detection methods. Furthermore, even provided with not purely “clean” data (unclean data), the proposed method is still robust and effective.

This paper presents a computational intelligence approach for predicting missing data in the presence of concept drift using an ensemble of multi-layered feed forward neural networks. An algorithm that detects concept drift by measuring heteroskedasticity is proposed. Six instances prior to the occurrence of missing data are used to approximate the missing values. The algorithm is applied to simulated time series data sets resembling non-stationary data from a sensor. Results show that the prediction of missing data in non-stationary time series data is possible but is still a challenge. For one test, up to 78% of the data could be predicted within 10% tolerance range of accuracy.

Radar high-resolution range profile (HRRP) has received intensive attention from the radar automatic recognition (RATR) community. Since the initial phase of a complex HRRP is strongly sensitive to target position variation, which is referred to as the initial phase sensitivity, only the amplitude information in the complex HRRP, what is called the real HRRP, is used for RATR. This paper proposes a novel feature extraction method for the complex HRRP. The extracted complex feature vector contains the difference phase information between range cells but no initial phase information in the complex HRRP. The recognition algorithms, frame-template-database establishment methods and preprocessing methods used in the real HRRP-based RATR can also be applied to the proposed complex feature vector-based RATR. The recognition experiments based on measured data show that the proposed complex feature vector can obtain better recognition performance than the real HRRP if only the cell interval parameters are proper.

In this paper, we propose a probabilistic approach to feature selection for multi-class text categorization. Specifically, we regard document class and occurrence of each feature as events, calculate the probability of occurrence of each feature by the theorem on the total probability and utilize the values as a ranking criterion. Experiments on Reuters-2000 collection show that the proposed method can yield better performance than information gain and

χ

-square, which are two well-known feature selection methods.

This paper presents zero-crossing-based feature extraction for the speech recognition using neck-microphones. One of the solutions in noise-robust speech recognition is using neck-microphones which are not affected by the environmental noises. However, neck-microphones distort the original voice signals significantly since they only capture the vibrations of vocal tracts. In this context, we consider a new method of enhancing speech features of neck-microphone signals using zero-crossings. Furthermore, for the improvement of zero-crossing features, we consider to use the statistics of two adjacent zero-crossing intervals, that is, the statistics of two samples referred to as the second order statistics. Through the simulation for speech recognition using the neck-microphone voice command system, we have shown that the suggested method provides the better performance than other approaches using conventional speech features.

In this paper, we present two novel memetic algorithms (MAs) for gene selection. Both are synergies of Genetic Algorithm (wrapper methods) and local search methods (filter methods) under a memetic framework. In particular, the first MA is a Wrapper-Filter Feature Selection Algorithm (WFFSA) fine-tunes the population of genetic algorithm (GA) solutions by adding or deleting features based on univariate feature filter ranking method. The second MA approach, Markov Blanket-Embedded Genetic Algorithm (MBEGA), fine-tunes the population of solutions by adding relevant features, removing redundant and/or irrelevant features using Markov blanket. Our empirical studies on synthetic and real world microarray dataset suggest that both memetic approaches select more suitable gene subset than the basic GA and at the same time outperforms GA in terms of classification predictions. While the classification accuracies between WFFSA and MBEGA are not significantly statistically different on most of the datasets considered, MBEGA is observed to converge to more compact gene subsets than WFFSA.

Automatic communication signal (e.g., FM signal) classification and identification focus on finding the fine feature contained in the almost approximate noisy communication signal comprehensively identifying the the same or different version of transmitters in modern electronic warfare. Direct use of HOS becomes unavailable for on-line application because of its huge computation time and memory space especially in the case of high frequency FM signal. This paper presents a novel view to improve the HOS analysis efficiency by sub-sampling while preserving the noise-contaminated fine feature and eliminating the random Gaussian noise. FM signal-related feature bispectra are also introduced to translate the 2-D feature matching pattern to a 1-D one applicable for an optimal adaptive k-means iterative RBF classifier. Computer simulations show that this novel feature bispectra outperform AIB and SB in terms of computation time and recognition rate for on-line steady FM signal recognition.

In the image registration process, there always exists rotation transformation. The ordinary methods such as NCC (Normalized Cross Correlation Algorithm), SD (Square Difference Algorithm), SSDA (Sequential Similarity Detection Algorithm), are not suitable for rotated image registration. In this paper, a method based on circular template, intensity distribution and SD is proposed for rotation image registration. Through the CPs (Control Points) got by the proposed method, transformation model and least square method, the rotation parameters are obtained. Experimental results verify its effectiveness. Compared with the existing feature-based approaches, it is easier to obtain CPs and needs no salient objects.