Advances in Neural Networks – ISNN 2015

In this paper, a novel T-S fuzzy model based adaptive synchronization scheme for nonlinear large-scale systems with uncertainties and time-delay is proposed. Based on the universal approximation property of T-S fuzzy model, a nonlinear large-scale system is established and fuzzy adaptive controllers are designed under Parallel Distributed Compensation (PDC) for overcoming the unknown uncertainties in systems and the time-delay in communication. Furthermore, under some certain condition, this synchronization scheme can be transformed into pinning synchronization control, which will indeed save much resource. Finally, a numerical simulation example is taken to show the effectiveness of the proposed adaptive synchronization scheme.

In this paper, the finite-time control problem for Markov systems with partly known transition probabilities and polytopic uncertainties is investigated. The main result provided is a sufficient conditions for finite-time stabilization via state feedback controller, and a simpler case without controller is also considered, based on switched quadratic Lyapunov function approach. All conditions are shown in the form of LMIs. An illustrative example is presented to demonstrate the result.

In this paper, the hybrid function projective synchronization of unknown Cohen-Grossberg neural networks with time delays and noise perturbation is investigated. A hybrid control scheme combining open-loop control and adaptive feedback control is designed to guarantee that the drive and response networks can be synchronized up to a scaling function matrix with parameter identification by utilizing the LaSalle-type invariance principle for stochastic differential equations. Finally, the corresponding numerical simulations are carried out to demonstrate the validity of the presented synchronization method.

In this brief, a neural dynamic surface control algorithm is proposed for three-phase pulse width modulation voltage source rectifier with the parametric variations. Neural networks are employed to approximate the uncertainties, including the parametric variations and the unknown load-resistance. The actual control laws are derived by using the dynamic surface control method. Furthermore, a linear tracking differentiator is introduced to replace the first-order filter to calculate the derivative of the virtual control law. Thus, the peaking phenomenon of the filter is suppressed during the initial phase. The system stability is analyzed by using the Lyapunov theory. Simulation results are provided to validate the efficacy of the proposed controller.

In this paper, a terminal reaching law based sliding mode control (SMC) method for load frequency control (LFC) is investigated in interconnected power systems in the presence of wind turbines and generation rate constraint (GRC). Neural networks are adopted to compensate the entire uncertainties. Simulation results show the validity and robustness of the presented method.

In this paper, a novel

Q

-learning based policy iteration adaptive dynamic programming (ADP) algorithm is developed to solve the optimal control problems for discrete-time nonlinear systems. The idea is to use a policy iteration ADP technique to construct the iterative control law which stabilizes the system and simultaneously minimizes the iterative

Q

function. Convergence property is analyzed to show that the iterative

Q

function is monotonically non-increasing and converges to the solution of the optimality equation. Finally, simulation results are presented to show the performance of the developed algorithm.

An adaptive neural network control approach is proposed for a class of stochastic nonlinear strict-feedback systems with unknown nonlinear function in this paper. Only one NN (neural network) approximator is used to tackle unknown nonlinear functions at the last step and only one actual control law and one adaptive law are contained in the designed controller. This approach simplifies the controller design and alleviates the computational burden. The Lyapunov Stability analysis given in this paper shows that the control law can guarantee the solution of the closed-loop system uniformly ultimate boundedness (UUB) in probability. The simulation example is given to illustrate the effectiveness of the proposed approach.

In this paper, the

H

 ∞ 

optimal control for a class of continuous-time nonlinear systems is investigated using event-triggered method. First, the

H

 ∞ 

optimal control problem is formulated as a two-player zero-sum differential game. Then, an adaptive triggering condition is derived for the closed loop system with an event-triggered control policy and a time-triggered disturbance policy. For implementation purpose, the event-triggered concurrent learning algorithm is proposed, where only one critic neural network is required. Finally, an illustrated example is provided to demonstrate the effectiveness of the proposed scheme.

In this paper, an observer-based adaptive control scheme for a class of nonlinear systems with parametric uncertainties is proposed. The adaptive observers using parameter estimates ensure the identification errors of system states are convergent to zero, and force the parameter estimates approach to the true values especially if the observer gains are selected large enough. By combining the Lyapunov synthesis with backstepping framework, the global asymptotical stability and bounded signals of the resulting closed-loop system can be ensured. A numerical example is employed to demonstrate the effectiveness of the proposed adaptive control scheme.

The

H

 ∞ 

control problem is considered for linear parabolic partial differential equation (PDE) systems with completely unknown system dynamics. We propose a model-free policy iteration (PI) method for learning the

H

 ∞ 

control policy by using measured system data without system model information. First, a finite-dimensional system of ordinary differential equation (ODE) is derived, which accurately describes the dominant dynamics of the parabolic PDE system. Based on the finite-dimensional ODE model, the

H

 ∞ 

control problem is reformulated, which is theoretically equivalent to solving an algebraic Riccati equation (ARE). To solve the ARE without system model information, we propose a least-square based model-free PI approach by using real system data. Finally, the simulation results demonstrate the effectiveness of the developed model-free PI method.

In this paper, intermittent control scheme is adopted to investigate the exponential synchronization of complex delayed dynamical networks with uncertain parameters. Based on Lyapunov function method and mathematical analysis technique, some novel and useful criteria for exponential synchronization are established. Finally, two numerical simulations are given to illustrate the effectiveness and correctness of the derived theoretical results.

With the superiority of owning more degrees of freedom than ordinary robot manipulators, redundant robot manipulators have gotten much attention in recent years. In order to control the trajectory of the robot end-effector with a desired velocity, it is very popular to apply the inverse kinematics approaches, such as pseudo-inverse scheme. However, calculating the inverse of Jacobian matrix requires a lot of time. Thus base on gradient neural dynamics (GND), an inverse-free scheme is proposed at the joint-velocity level. The scheme is named G1 type as it uses GND once. In addition, two path tracking simulations based on five-link and six-link redundant robot manipulators illustrate the efficiency and the accuracy of the proposed scheme. What is more, the physical realizability of G1 type scheme is also verified by a physical experiment based on the six-link planar redundant robot manipulator hardware system.

This paper presents a closed-loop trajectory tracking controller for an Unmanned Underwater Vehicle(UUV) with five degrees of freedom. A backstepping control (BSC) methodology combined with Lyapunov theorem is adopted to design the controller of trajectory tracking. Then an online-tuning fuzzy neural network (FNN) framework is chosen to inherit the conventional BSC law. Moreover, the adaptive parameters tuning laws are derived in the sense of Lyapunov stability theorem and projection algorithm to ensure the network convergence as well as stable control performance. Finally, the simulation results on UUV verify that an excellent performance of the proposed controller can be obtained.

This paper is concerned with the sampled-data state estimation problem for neural networks of neutral-type with time-varying delays. A new state estimator was designed based on the sampled measurements. The sufficient condition for the existence of state estimator is derived by using the Lyapunov functional method. A numerical example is given to show the effectiveness of the proposed estimator.

In this paper, we investigate the exponential lag synchronization of delayed Cohen-Grossberg neural networks with discontinuous activation functions. By employing the analysis technique and theory of the differential equations with discontinuous right-hand side, some novel lag synchronization criteria have been obtained. Finally, an example is given to illustrate the effectiveness of the obtained results.

This paper is concerned with globally exponential stability in the mean square of stochastic static neural networks with Markovian switching and time delay. Firstly, the mathematical model of this kind of recurrent neural networks is established by taking information latching and noise disturbance into consideration. Then, a stability condition, which is dependent on both time delay and system mode, is presented in terms of linear matrix inequalities. Based on it, the maximum value of the exponential decay rate can be efficiently found by solving a convex optimization problem.

In this paper, we are concerned with the robust multistability and multiperiodicity of delayed neural networks. A set of sufficient conditions ensuring the coexistence of 2

n

periodic solutions and their local stability are presented. And the attraction basin of each periodic solution can be enlarged by rigorous analysis.

A novel memristor-based hyperchaotic system is proposed and studied in this paper. The memristor is nonlinear memory element intrinsically, which has the potential application for generating complex dynamics in nonlinear circuit to reduce system power consumption and circuit size. As the non-linear part of a system, the HP memristor is introduced to a four-dimensional system. Chaotic attractors, Lyapunov exponent spectrum, Lyapunov dimension, power spectrum, Poincaré map and bifurcation with respect to various circuit parameter, are considered and observed, which together demonstrate the rich chaotic dynamical behaviors of the system. Finally, the circuit in SPICE are designed for the proposed memristive hyperchaotic system. The SPICE experimental results are consistent with the numerical simulation results, which verifies the feasibility of the memristor hyperchaotic system.

This letter uses image overlay technique on memristor crossbar array (MCA) structure for image storing. Different programming circuits with time slot techniques are designed for the MCA consisting of the nonlinear HP memristor (HPMCA) and the MCA composed of the piece-wise linear threshold memristor (TMCA). The experiment results indicate that the HPMCA has a better performance, the TMCA is more practical in the industrial implementation. As a conclusion, the MCA made up of the memristor with both the nonlinear drift boundary property and the threshold property is preferred for image overlay.

In this paper, we study the global exponential stability in Lagrange sense for memristor-based neural networks (MBNNs) with time-varying delays. Based on the nonsmooth analysis and differential inclusion theory, matrix measure technique is employed to establish some succinct criteria which ensure the Lagrange stability of the considered memristive model. In addition, the new proposed criteria are very easy to verify, and they also enrich and improve the earlier publications. Finally, two example are given to demonstrate the validity of the results.

In this paper, a general class of non-monotonic piecewise linear activation functions is introduced and then the coexistence and dynamical behaviors of multiple equilibrium points are studied for a class of memristive neural networks (MNNs). It is proven that under some conditions, such

n

-neuron MNNs can have 5

n

equilibrium points located in

$\Re^n$

, and 3

n

of them are locally exponentially stable, by means of fixed point theorem, nonsmooth analysis theory and rigorous mathematical analysis. The investigation shows that the neural networks with non-monotonic piecewise linear activation functions introduced in this paper can have greater storage capacity than the ones with Mexican-hat-type activation function.

This paper investigates the problem of global exponential anti-synchronization of a class of memristive chaotic neural networks with time-varying delays. First, a memrsitive neural network is modeled. Then, considering the state-dependent properties of the memristor, a new fuzzy model employing parallel distributed compensation (PDC) provides a new way to analyze the complicated memristive neural networks with only two subsystems. And the controller is dependent on the output of the system in the case of packed circuits. An illustrative example is also presented to show the effectiveness of the results.

In this paper, we develop an interactive Non-negative Matrix Factorization method for representative action video discovery. The original video is first evenly segmented into some short clips and the bag-of-words model is used to describe each clip. Then a temporal consistent Non-negative Matrix Factorization model is used for clustering and action segmentation. Since the clustering and segmentation results may not satisfy the user’s intention, two extra human operations: MERGE and ADD are developed to permit user to improve the results. The newly developed interactive Non-negative Matrix Factorization method can therefore generate personalized results. Experimental results on the public Weizman dataset demonstrate that our approach is able to improve the action discovery and segmentation results.

In this paper, a new method for image texture representation is proposed, which represents image content using a 49 dimensional feature vector through calculating the variation of texture direction and the intensity of texture. In addition, the texture feature is grouped into a feature set with some other image texture representation methods, and then a new online feature selection method with a novel discrimination criterion is presented. We test the discriminating ability of every feature in the feature set utilizing the discrimination criterion, and select the optimal feature subset, which expresses image content in an even better fashion. The results of the computer simulation experiments show that the proposed feature extraction and feature selection method can represent image content effectively, and improve the retrieval precision visibly.

Face parsing is a basic task in face image analysis. It amounts to labeling each pixel with appropriate facial parts such as eyes and nose. In the paper, we present a interlinked convolutional neural network (iCNN) for solving this problem in an end-to-end fashion. It consists of multiple convolutional neural networks (CNNs) taking input in different scales. A special interlinking layer is designed to allow the CNNs to exchange information, enabling them to integrate local and contextual information efficiently. The hallmark of iCNN is the extensive use of downsampling and upsampling in the interlinking layers, while traditional CNNs usually uses downsampling only. A two-stage pipeline is proposed for face parsing and both stages use iCNN. The first stage localizes facial parts in the size-reduced image and the second stage labels the pixels in the identified facial parts in the original image. On a benchmark dataset we have obtained better results than the state-of-the-art methods.

The problem of tag completion is to learn the missing tags of an image. In this paper, we propose to learn a tag scoring vector for each image by local linear learning. A local linear function is used in the neighborhood of each image to predict the tag scoring vectors of its neighboring images. We construct a unified objective function for the learning of both tag scoring vectors and local linear function parameters. In the objective, we impose the learned tag scoring vectors to be consistent with the known associations to the tags of each image, and also minimize the prediction error of each local linear function, while reducing the complexity of each local function. The objective function is optimized by an alternate optimization strategy and gradient descent methods in an iterative algorithm. We compare the proposed algorithm against different state-of-the-art tag completion methods, and the results show its advantages.

Haarlike feature has achieved great success in detecting frontal human faces, but fewer attentions have been paid to the other objects such as pedestrian. The reason of the low detection rate for Haarlike feature is attributed to the usage in a naive way. In this paper, we have revisited Haarlike feature for object detection especially focus on pedestrians, but use it in a different way which is applied based on multiple channel maps instead of raw pixels and obtains a significant improvement. Furthermore, we have proposed an improved Haarlike feature that embeds statistical information from the training data which is based on the linear discriminative analysis criterion. The proposed feature works with the classical Gentle Boosting algorithm which is effective in training, and also running at real-time speed. Experiments based on INIRA dataset demonstrate that our proposed method is easy to implement and achieves the performance comparable to the state-of-the-arts.

Computer vision methods can benefit wood processing industry. We propose a method to detect wood surface quality and classify wood samples into sound and defective classes. Gray level histogram statistical features and gray level co-occurrence matrix (GLCM) texture features are extracted from wood surface images and combined for classification. A half circle template is proposed to generate GLCM, avoiding calculating distances at each pixel every time and speeding up the algorithm greatly. The proposed approach uses more pixel information than traditional four-angle method, resulting in a significantly higher classification accuracy. Moreover the running time demonstrates our algorithm is efficient and suitable for real-time applications.

A robust satellite image classification is the fundamental step for aerial image understanding. However current methods with hand-crafted features and conventional classifiers have limited performance. In this paper we introduced convolutional neural network (CNN) method into this problem. Two approaches, including using conventional classifier with CNN features and direct classification with trained CNN models, are investigated with experiments. Our method achieved 97.4% accuracy on 5-fold cross-validation test of the UCMERCED LULC dataset, which is 8% higher than state-of-the-art methods.

Based on the different strength of synaptic connections between actual neurons, this paper proposes a novel heterogeneous PCNN (HPCNN) algorithm to quantize images. HPCNN is constructed with traditional pulse coupled neural network (PCNN) models, which has different parameters corresponding to different image regions. It puts pixels of different gray levels to be classified broadly into two categories: the background regional ones and the object regional ones. Moreover, HPCNN also satisfies human visual characteristics (HVS). The parameters of HPCNN model are calculated automatically according to these categories and quantized results will be optimal and more suitable for human to observe. At the same time, the experimental results show the validity and efficiency of our proposed quantization method.

Image fusion algorithm is a key technology to eliminate noise through combining each image with different weight. Recently, convergence and convergence speed are two exiting problems which attract more and more attention. In this paper, we originally propose a image fusion algorithm based on neural network. Firstly, the linearly constrained least square(LCLS) model which can deal with image fusion problem is introduced. In addition, in order to handle LCLS model, we adopt the penalty function technique to construct a neural network. The proposed algorithm has a simpler structure and faster convergence speed. Lastly, simulation results show this fusion algorithm which has great ability to remove different noise.

In this paper, a novel singing voice/music separation method is proposed based on the non-negative tensor factorization (NTF) and repeat pattern extraction technique (REPET) to separate the mixture into an audio signal and a background music. Our system consists of three stages. Firstly, we use the NTF to decompose the mixture into different components, and similarity detection is applied to distinguish the components from each other, in order to classify the components into two classes as the voice including voice/periodic music and the block music/voice; next we utilize the REPET to extract the background music one step further for the two classes, and the final background music is estimated by adding the two backgrounds together, the left is added together as the singing voice; finally the music spectrum and the voice spectrum are filtered by harmonic filter and percussive filter respectively. To improve the performance further, wiener filter is used to separate the voice and music. Our method can improve the separation performance compared with the other state-of-the-art methods on the MIR-1K dataset.

Cervical vertebrae are important ramus communican that connect human body and the corpus. Muscles around cervical vertebrae such as deep cervical flexor and sternocleidomastoid muscle do key role to control chronicle neck pain thus monitoring such muscles near cervical vertebrae is important. In this paper, we propose a method to detect and analyze cervical vertebrae and related muscles automatically with fuzzy ART clustering from ultrasonography. The experiment verifies that our approach is consistent with human medical experts’ decision to locate key measuring point for muscle analysis and successful in detecting cervical vertebrae accurately.

The adaptive Fourier decomposition (AFD) is a greedy iterative signal decomposition algorithm in the viewpoint of energy. Instead of using a fixed basis for decomposition, AFD uses an adaptive basis to achieve efficient energy extraction. In the conventional searching method, a new basis is searched from a large dictionary at every decomposition level. This usually results in a slow searching speed. To improve the efficiency, a fast searching method based on Nelder-Mead algorithm is proposed in this paper. The AFD with the proposed searching method is applied for electrocardiography (ECG) signals in which the selection ranges of four key parameters in the proposed searching method are determined based on simulation results of an artificial ECG signal. The simulation results of real ECG data shows that the computational time of the AFD based on the proposed searching method is just half of that based on the conventional searching method with similar reconstruction error.

Among the EEG-based BCIs, SSVEP-based BCIs have gained much attention due to the advantages of relatively high information transfer rate (ITR) and short calibration time. Although in SSVEP-based BCIs the frequency recognition methods using multiple channels EEG signals may provide better accuracy, using single channel would be preferable in a practical scenario since it can make the system simple and easy-to-use. To this goal, we propose a new single channel method based on wavelet-independent component analysis (WICA) in the SSVEP-based BCI, in which wavelet transform (WT) is applied to decompose a single channel signal into several wavelet components and then independent component analysis (ICA) is applied to separate the independent sources from the wavelet components. Experimental results show that most of the time the recognition accuracy of the proposed single channel method is higher than the conventional single channel method, power spectrum (PS) method.

The mixture of Gaussian processes (MGP) is a powerful statistical learning model for regression and prediction and the EM algorithm is an effective method for its parameter learning or estimation. However, the feasible EM algorithms for MGPs are certain approximations of the real EM algorithm since Q-function cannot be computed efficiently in this situation. To overcome this problem, we propose an MCMC based EM algorithm for MGPs where Q-function is alternatively estimated on a set of simulated samples via the Markov Chain Monte Carlo (MCMC) method. It is demonstrated by the experiments on both the synthetic and real-world datasets that our proposed MCMC based EM algorithm is more effective than the other three EM algorithms for MGPs.

For the learning of mixtures of Gaussian processes, model selection is an important but difficult problem. In this paper, we develop an automatic model selection algorithm for mixtures of Gaussian processes in the light of the reversible jump Markov chain Monte Carlo framework for Gaussian mixtures. In this way, the component number and the parameters are updated according the five types of random moves and model selection can be made automatically. The key idea is that the moves of component splitting or merging preserve the zeroth, first and second moments of the components so that the covariance parameters of the new components can be related to the origin ones. It is demonstrated by the simulation experiments that this automatic model selection algorithm is feasible and effective.

The Mixture of Gaussian Processes (MGP) is a powerful statistical learning framework in machine learning. For the learning of MGP on a given dataset, it is necessary to solve the model selection problem, i.e., to determine the number

C

of actual GP components in the mixture. However, the current learning algorithms for MGPs cannot solve this problem effectively. In this paper, we propose an effective model selection criterion, called the Synchronously Balancing or SB criterion for MGPs. It is demonstrated by the experimental results that this SB criterion is feasible and even outperforms two classical criterions: AIC and BIC, for model selection on MGPs. Moreover, it is found that there exists a feasible interval of the penalty coefficient for correct model selection.

A new algorithm for single hidden layer feedforward neural networks (SLFN), Orthogonal Basis Extreme Learning (OBEL) algorithm, is proposed and the algorithm derivation is given in the paper. The algorithm can decide both the NNs parameters and the neuron number of hidden layer(s) during training while providing extreme fast learning speed. It will provide a practical way to develop NNs. The simulation results of function approximation showed that the algorithm is effective and feasible with good accuracy and adaptability.

Text clustering is an important research topic in data mining. Many text clustering methods have been proposed and obtained satisfactory results. Information Bottleneck algorithm, which is based on information loss, can measure complicated relationship between variables. It is taken as one of the most informative text clustering methods and has been applied widely in practical. With the development of information technology, the scale of text becomes larger and larger. Classical information bottleneck based clustering method will be out of work to process large-scale dataset because of expensive computational cost. For dealing with large scale text clustering problem, a novel clustering method based on MapReduce is proposed. In the method, dataset is divided into sub datasets and deployed to different computational nodes. Each computational node will only process sub dataset. The computational cost can be reduced markedly. The efficiency of the method is illustrated with a practical text clustering problem.

In this paper, we study the problem of using contextual data points of a data point for its classification problem. We propose to represent a data point as the sparse linear reconstruction of its context, and learn the sparse context to gather with a linear classifier in a supervised way to increase its discriminative ability. We proposed a novel formulation for context learning, by modeling the learning of context reconstruction coefficients and classifier in a unified objective. In this objective, the reconstruction error is minimized and the coefficient sparsity is encouraged. Moreover, the hinge loss of the classifier is minimized and the complexity of the classifier is reduced. This objective is optimized by an alternative strategy in an iterative algorithm. Experiments on three benchmark data set show its advantage over state-of-the-art context-based data representation and classification methods.

In this paper, a novel K-means evolving spiking neural network (K-ESNN) model for clustering problems has been presented. K-means has been utilised to improve the original ESNN model. This model enhances the flexibility of the ESNN algorithm in producing better solutions to overcoming the disadvantages of K-means. Several standard data sets from UCI machine learning are used for evaluating the performance of this model. It has been found that the K-ESNN provides competitive results in clustering accuracy and speed performance measures compared to the standard K-means. More discussion is provided to prove the effectiveness of the new model in clustering problems.

Individuals in large groups respond to the movements and positions of their neighbors by following a set of interaction rules. These rules are central to understanding the mechanisms of collective motion. However, whether individuals actually use these rules to guide their movements remains untested. Here we show that the real-time movements of individual fish can be directly predicted from their neighbors’ motion. We train a recurrent neural network to predict the trajectories of individual fish from input signals. The inputs are projected to the recurrent network as time series representing the movements and positions of neighboring fish. By comparing the data output from the model with the target fish’s trajectory, we provide direct evidence that individuals guide their movements via interaction rules. Because the error between the model output and actual trajectory changes when the fish perceive a noxious contaminant, the model is potentially applicable to water quality monitoring.

Reliable and accurate short-term wind speed forecasting is of great importance for secure power system operations. In this study, a novel two-step method to construct a multi-model ensemble, which consists of linear regression, multi-layer perceptrons and support vector machines, is proposed. The ensemble members first compete with each other in a number of training rounds, and the one with the best forecasting accuracy in each round is recorded. Then, after all the training rounds, the occurrence frequency of each member is calculated and used as the weight to form the final multi-model ensemble. The effectiveness of the proposed multi-model ensemble has been assessed on the real datasets collected from three wind farms in China. The experimental results indicate that the proposed ensemble is capable of providing better performance than the single predictive models composing it.

Orthogonal frequency division multiplexing (OFDM) introduces large peak power of transmitted signals in time, which can result in significant signal distortion in the presence of nonlinear amplifiers. Partial transmit sequence (PTS) are well-known techniques for peak-power reduction in OFDM. However, the exhaustive search of phase factors in conventional PTS causes high computational complexity. In this paper, we present a suboptimal strategy for combining partial transmitted sequences that achieve good balance between computational complexity and power control performance. The simulation results show that the proposed algorithm can not only reduces the PAPR significantly, but also decreases the computational complexity

This paper presents new results on neurodynamic optimization approach to solve bilevel linear programming problems (BLPPs) with linear inequality constraints. A sub-gradient recurrent neural network is proposed for solving the BLPPs. It is proved that the state convergence time period is finite and can be quantitatively estimated. Compared with existing recurrent neural networks for BLPPs, the proposed neural network does not have any design parameter and can solve the BLPPs in finite time. Some numerical examples are introduced to show the effectiveness of the proposed neural network.

In this paper, the stability of a novel nonlinear neural network solving linear programming problems is studied. We prove that this nonlinear neural network is stable in the sense of Lyapunov under certain conditions. Inspired by the study of this neural network, we propose a novel neural system to solving the

k

-winners-take-all (

k

WTA) problem. Numerical simulations demonstrate that the effectiveness and good performance of our new

k

WTA neural network.

This paper presents a continuous-time multi-agent network for distributed least absolute deviation (DLAD). The objective function of the DLAD problem is a sum of many least absolute deviation functions. In the multi-agent network, each agent connects with its neighbors locally and they cooperate to obtain the optimal solutions with consensus. The proposed multi-agent network is in fact a collective system with each agent being considered as a recurrent neural network. Simulation results on a numerical example are presented to illustrate the effectiveness and characteristics of the proposed distributed optimization method.

In this paper, online solution of complex-valued systems of linear equations is investigated in the complex domain. Different from the conventional real-valued neural network, which is only designed for real-valued linear equations solving, a fully complex-valued gradient neural network (GNN) is developed for online complex-valued systems of linear equations. The advantages of the proposed complex-valued GNN model decrease the unnecessary complexities in theoretical analysis, real-time computation and related applications. In addition, the theoretical analysis of the fully complex-valued GNN model is presented. Finally, simulative results substantiate the effectiveness of the fully complex-valued GNN model for online solution of the complex-valued systems of linear equations in the complex domain.

Sparse representation in sensory cortex has been well verified and its capability of yielding response properties of single neurons is also demonstrated. In order to improve sparse representation to be more neurally plausible, we reconsider several response properties of single neurons, especially the cross orientation suppression and surround suppression. A new sparse representation model using intracellular and extracellular neural mechanisms is presented. Simulation results of the presented model explain physiological observations very well.

Some load balancing algorithms in heterogeneous wireless networks can not consider the problems arising from the admission control of new service and service transfer of heavy load networks. To solve these problems, we propose a load balancing algorithm based on neural networks. This algorithm is used to conduct prediction through network load rate and achieve the network admission of new service by combining an admission control optimization algorithm. Moreover, by analyzing network performance, some services of heavy load network are transferred to overlay light load network. The simulation results indicate that our algorithm can well realize the load balancing of heterogeneous wireless network and provide high resource utilization.

In this paper, kinds of network applications are first analyzed, and some simple and effective features from the package headers of network flows are then generated by using the method of time window. What is more, three kinds of machine learning algorithms, which are support vector machine (SVM), back propagation (BP) neural network and BP neural network optimized by particle swarm optimization (PSO), are developed respectively for training and identification of network traffic. The experimental results show that traffic identification based on SVM can not only quickly generate classifier model, but also reach the accuracy of more than 98% under the condition of small sample. Moreover, the method proposed by this paper can measure and identify Internet traffic at any time and meet the needs of identifying real-time multi-application.

Indeed, countermeasures, as well as computer viruses, could spread in the network. This paper aims to investigate the effect of propagation of countermeasures on viral spread. For the purpose, a new virus-antivirus spreading model is proposed. The global asymptotic stability of the virus-free equilibrium is proved when the threshold is below the unity, and the existence of the viral equilibrium is shown when the threshold exceeds the unity. The influences of different model parameters on the threshold are also analyzed. Numerical simulations imply that the propagation of countermeasures contributes to the suppress of viruses, which is consistent with the fact.

A huge increase in the number of mobile malware brings a serious threat to Internet security, as the adoption rate of mobile device is soaring, especially Android device. A variety of researches have been developed to defense malware, but the mobile device users continuously suffer private information leak or economic losses from malware. Recently, a large number of methods have been proposed based on static or dynamic features analysis combining with machine learning methods, which are considered effective to detect malware on mobile device. In this paper, we propose an effective framework to detect malware on Android device based on feature extraction and neural network calssifier. In this framework, we take use of static features to represent malware and utilize extreme learning machine (ELM) algorithm to learn the neural network. We first extract features from the malware, and then utilize three different feature extraction methods including principal component analysis (PCA), Karhunen-Loève transform (KLT) and independent component analysis (ICA) to transform the feature matrix into new feature spaces and generate three new feature matrixes. For each feature matrix, we construct

En

base classifiers by using ELM. Finally, we utilize Stacking method to combine the results. Experimental results suggest that the proposed framework is effective in detecting malware on Android device.

In modeling, prediction and control applications, the single-input-rule-modules (SIRMs) connected fuzzy inference method can efficiently tackle the rule explosion problem that conventional fuzzy systems always face. In this paper, to improve the learning performance of the SIRMs method, a neural structure is presented. Then, based on the least square method, a novel parameter learning algorithm is proposed for the optimization of the SIRMs connected neural-fuzzy system. Further, the proposed neural-fuzzy system is applied to the cooling and heating loads prediction which is a popular multi-variable problem in the research domain of intelligent buildings. Simulation and comparison results are also given to demonstrate the effectiveness and superiority of the proposed method.