Search

Sensitive feature extraction from the raw vibration signal is still a great challenge for intelligent fault diagnosis of rolling bearing. Current fault classification framework generally concentrates on the pattern of classifier with single-domain feature, which is easy to induce insufficient feature extraction and low recognition accuracy. Therefore, to address this issue and improve intelligent diagnostic accuracy of rolling bearing, this paper proposes a novel fault classification algorithm based on optimized SVM with multi-domain feature, which mainly consists of three stages (i.e. multi-domain feature extraction, feature selection and fault identification). In this first stage, three approaches (i.e. statistical analysis, FFT and VMD) are separately applied to extract the fault feature information from multi-domain aspect (e.g. time-domain, frequency-domain and time-frequency domain), which can excavate comprehensively the condition information and intrinsic property of the raw vibration signal. Secondly, Laplace score algorithm is introduced to select automatically the meaningful sensitive feature according to the importance of each feature, which is aimed at removing some redundant information and improving the calculation efficiency. Finally, particle swarm optimization-based support vector machine (PSO-SVM) classification model is employed to implement the identification of multiple fault condition of rolling bearing. Performance of the proposed method is evaluated on two experimental examples of rolling bearing fault diagnosis. Experimental results show that the proposed method achieves high diagnosis accuracy for different working conditions of rolling bearing and outperforms some traditional methods both mentioned in this paper and published in other literature.

In this paper, synchronization of stochastic complex networks (SCNs) with time delay is researched. A feedback control method is introduced to drive such SCNs to achieve inner synchronization in mean square. Different from previous researches, the feedback control we designed is based on discrete-time state observations. Moreover, by exploiting Lyapunov method and Kirchhoff’s Matrix Tree Theorem in graph theory, some sufficient criteria are obtained to guarantee synchronization in mean square and asymptotical synchronization in mean square of SCNs. What is more, we make use of the theoretical results to analyze asymptotical synchronization in mean square of second-order Kuramoto oscillators with time delay and stochastic disturbances and get a sufficient criterion. At the end of this paper, a numerical example is provided to validate the effectiveness and feasibility of our theoretical results.

Multi-label classification has attracted great attention from different domains. RAkEL (Random k-label sets) is an effective high-order multi-label learning approach. However, this method exploits label correlations randomly, which cannot make full use of label correlations, causing poor performance of sub-classifiers. In the paper, a novel ensemble method is proposed for multi-label classification. Instead of randomly selecting small-size subsets, this paper constructs a set of k-labelset based on local positive and negative pairwise label correlations, which are captured by KNN and matrix similarity measure. Furthermore, considering the possible missing or noisy labels in predictions, a rectification mechanism based on local label correlations is established to lessen global errors, which enhances the accuracy of the classification procedure. The experimental results over eight benchmark datasets based on multiple evaluation metrics show that these two strategies can achieve relatively better performance compared with the RAkEL method and other state-of-the-art approaches.

This paper investigates the disturbance observer-based adaptive neural tracking control of a class of multiple-input multiple-output (MIMO) systems in the presence of unmodeled dynamics, system uncertainties, time varying disturbance and input dead-zone. An adaptive neural control method combined with backstepping technique and the radial basis function neural networks (RBFNNs) is proposed for the systems under consideration. In recursive backstepping designs, neural network (NN) is employed for uncertainty approximation. The disturbance observer is developed to provide efficient learning of the compounded disturbance which includes the effect of time varying disturbance, neural network approximation error. It is shown that by using Lyapunov methods, the developed control scheme can ensure semi-globally uniformly ultimately bounded (SGUUB) of all signals within the closed-loop systems. Simulation results are presented to illustrate the validity of the approach. This paper is novel at least in the two aspects: (1) disturbance observer based tracking control method is developed for MIMO nonlinear systems with unmodeled dynamics and (2) the strong coupled terms is considered in this paper where the interconnections are functions of all states, which is a more general form than existing related results.

In this paper, we are concerned with the finite-time synchronization for a class of complex-valued neural networks with time delays. Instead of using some finite-time stability theorems which are recently widely applied to investigating the finite-time synchronization for neural networks, by means of using integral inequality method, two novel sufficient conditions on the finite-time synchronization for the above delayed complex-valued neural networks are established. Our results and method on finite-time synchronization for the above neural networks are new and complementary to the existing papers.

Clustering analysis is to classify elements into categories based on their similarity. Clustering by fast search and find of density peaks (CFSFDP) has been proven to be an effective and novel algorithm, which identifies the centers of clusters with density maxima. However, the performance of CFSFDP is quite sensitive to the estimation of densities, that is exactly the selection of the cutoff distance (dc). In a conventional way, the selection of dc is based on subjective experience. It meets difficulties in finding an appropriate dc, especially for detecting nonspherical clusters, because CFSFDP cannot perform well when there are more than one density peak for one cluster. Besides, another barrier of applying CFSFDP is that manual interaction is always required for making an effective selection of cluster centers. In this paper, a new density-based clustering algorithm, clustering by defining and merging candidates of cluster centers via independence and affinity (CDMC-IA), is proposed. With its strategy, an appropriate value of cutoff distance dc can be well suggested and the robustness of the method itself is enhanced. Moreover, CDMC-IA introduces a new quantity independence to sort and select cluster centers, instead of human based selection from decision graph. Another quantity affinity is also introduced, which well handles multiple density peaks existing in one cluster and is able to assign each data point to the its targeted cluster. The performance of applying conventional clustering methods to benchmark datasets will be compared with the proposed method in this paper.

Feature selection is an essential pre-processing part in multi-label learning. Multi-label learning is usually used to deal with many complicated tasks, in which each sample is associated with multiple labels simultaneously. Fuzzy rough set model is one of the most effective ways for multi-label learning. However, it treats feature space and label space separately, and only uses features to describe sample structure information. In this paper, we fully consider the internal correlation between feature space and label space while fusing kernelized information from respective spaces. Moreover, we integrate fuzzy rough set with multiple kernel learning to finally realize feature selection. To be specific, firstly, we leverage one kind of kernel function to reveal the similarity between samples in feature space, and another one to assess the degree of label overlap between samples in label space. Secondly, we combine the kernelized information from the two spaces through linear combination to achieve precisely the lower approximation and construct a robust multi-label kernelized fuzzy rough set model, called RMFRS in this paper. Meanwhile, we discuss its properties and give theoretical analysis. Finally, we define a measurement criterion for selecting optimal features to evaluate the performance of the proposed algorithm. As many as 10 publicly available data sets are used to validate the effectiveness of our methods, and the result shows a distinct advantage over the state-of-the-art.

Decomposition methods play an important role in solving large-scale quadratic programming (QP) problems arising from support vector machines (SVMs). In this paper, we study convergence of general decomposition methods for SVMs. We prove that, under a mild condition on the working set selection, a decomposition algorithm stops within a finite number of iterations after reaching a solution of the QP problem satisfying a relaxed Karush–Kuhn–Tucker (KKT) condition which has been often used so far. Further, it is shown that the working set selection used in the implementation of SVMlight satisfies the condition given in this paper, so SVMlight has the finite termination property without the stronger assumption than the positive-semi-definiteness on the Hessian matrix of the objection function.

This paper deals with the model-free optimal containment control problem for a class of linear multi-agent systems (MASs). In the existing results concerning containment control of MASs, the dynamics of the MASs is required to be completely known. Differently, in this paper, we propose a new distributed self-learning control scheme based on action dependent heuristic dynamic programming (ADHDP) to achieve containment control, where the model of MASs is no longer needed. The containment control problem is first transformed into a regulation problem on the dynamics of the designed local containment error. The policy iteration method based on the designed state-action value function (also called the Q-function) is introduced to deal with such a regulation problem. The convergence analysis of this policy iteration method is also given. Neural network (NN) based actor-critic framework is adopted to approximate the optimal Q-functions and the optimal control policies for facilitating the implementation of the proposed method. It shows that the approximated control policies achieve the containment control and satisfy the global Nash equilibrium. Finally, the simulation results are provided to demonstrate the effectiveness of the developed approach.

This paper is concerned with the H∞ filtering problem for discrete-time interconnected fuzzy systems with partially unknown membership functions and past output measurements. A new H∞ filter with past output measurements is designed by introducing a switching mechanism which depends on the lower and upper bounds of the unknown elements of the membership functions, and graph theory is applied to addressed the interconnections, such that the filtering error system is asymptotically stable and the desired performance is guaranteed. By using Lyapunov analysis method, H∞ filter design conditions are given in terms of convex problem constrained by linear matrix inequalities (LMI). The results of this paper extend the existing filter design methods and obtain less conservative conditions. Finally, a numerical example is given to illustrate the advantages of the proposed design method.

Short term electric load forecasting, as an important tool in the electricity market, plays a critical role in the management of electric systems. Proposing an accuracy and optimization method is not only a challenging task but also an indispensable part of the energy system. More and more accurate forecasting methods are needed by different people in different areas. This paper proposes a novel short-term electric load forecasting method EMD-Mixed-ELM which based on empirical mode decomposition (EMD) and extreme learning machine (ELM). EMD-Mixed-ELM first uses the empirical mode decomposition to decompose the load series for capturing the complicated features of the electric load and de-noising the data. Considering that the performance of extreme learning machine (ELM) is greatly influenced by the choice of kernel, the mixed kernel method is proposed for ELM. The mixed kernel combines the RBF kernel and the UKF kernel. The forecasting results of the EMD-Mixed-ELM are proved to be better than all the other three methods (RBF-ELM, UKF-ELM and Mixed-ELM) and other existing methods (MFES, ESPLSSVM and Combined method). To verify the forecasting ability of the EMD-Mixed-ELM, half-hourly electric load data from the state of New South Wales, Victoria and Queensland in Australia are used in this paper as a case study. The experimental results clearly indicate that for this three datasets, the forecasting accuracy of the proposed method is superior to other methods.

Image captioning means automatically generating a caption for an image. As a recently emerged research area, it is attracting more and more attention. To achieve the goal of image captioning, semantic information of images needs to be captured and expressed in natural languages. Connecting both research communities of computer vision and natural language processing, image captioning is a quite challenging task. Various approaches have been proposed to solve this problem. In this paper, we present a survey on advances in image captioning research. Based on the technique adopted, we classify image captioning approaches into different categories. Representative methods in each category are summarized, and their strengths and limitations are talked about. In this paper, we first discuss methods used in early work which are mainly retrieval and template based. Then, we focus our main attention on neural network based methods, which give state of the art results. Neural network based methods are further divided into subcategories based on the specific framework they use. Each subcategory of neural network based methods are discussed in detail. After that, state of the art methods are compared on benchmark datasets. Following that, discussions on future research directions are presented.

This paper focuses on the task of targeted sentiment classification over dynamic memory networks. To remedy the flaw of the previous related work, only doing well in simple sentences, the new dynamic memory networks, in which the input module, question module and memory module are optimally designed, are established to model the task into question answering system for the first time. Specifically, in the input module, position embedding is considered to eliminate the influence of sentiment-irrelevant words that close to the target, and skip connection can help to build rich representations over the sentences. In the question module, for dealing with many cases where the target is composed of multiple words, the commonly used average target vector is replaced by the designed target sentiment question, which is encoded by a GRU. In the memory module, two improvements are performed. On the one hand, weight bias, caused by using original soft attention to extract memory information at each attention step, is avoided by both exploring attention and inner attention based GRU networks. On the other hand, memory updating block is capable to distill sentiment-relevant features from memory information, which is extracted by multiple attention block using multiple attention mechanism. Finally, experimental results demonstrate the effectiveness of the proposed models for the task of targeted sentiment classification among three different datasets. Additionally, the best model of this paper achieves state-of-the-art performance on different types of data.

This paper investigates the matrix projective synchronization (MPS) problem for time-varying disturbed complex dynamical networks (CDNs) with uncertain nonlinear structures and different dimensional nodes. In order to well describe the practical networks, the unavoidable unknown nonlinear structure and the inevitable uncertain disturbance of each node are considered into our network model, which is different from the previous works. Besides, it is worth pointing out that the outer coupling configuration matrix, which represents the coupling strength and the topological structure, is not restricted by the dissipatively coupled condition in this paper. The definition of MPS is also introduced for the networks with different dimensional nodes. Moreover, several MPS schemes are respectively put forward for our network model according to the norm bound of the uncertain nonlinear structures and disturbances being unknown or not. Finally, two proper examples associated with numerical simulations are given to verify the effectiveness of our theoretical results.

This work addresses the coordination problem of multiple robots with the goal of finding specific hazardous targets in an unknown area and dealing with them cooperatively. The desired behaviour for the robotic system entails multiple requirements, which may also be conflicting. The paper presents the problem as a constrained bi-objective optimization problem in which mobile robots must perform two specific tasks of exploration and at same time cooperation and coordination for disarming the hazardous targets. These objectives are opposed goals, in which one may be favored, but only at the expense of the other. Therefore, a good trade-off must be found. For this purpose, a nature-inspired approach and an analytical mathematical model to solve this problem considering a single equivalent weighted objective function are presented. The results of proposed coordination model, simulated in a two dimensional terrain, are showed in order to assess the behaviour of the proposed solution to tackle this problem. We have analyzed the performance of the approach and the influence of the weights of the objective function under different conditions: static and dynamic. In this latter situation, the robots may fail under the stringent limited budget of energy or for hazardous events. The paper concludes with a critical discussion of the experimental results., Comment: 40 pages

Control of blast furnace (BF) ironmaking process has always been a hot and difficult issue in metallurgic engineering and automation. In this paper, a novel data-driven inverse system identification based predictive control method is proposed for multivariate molten iron quality (MIQ) indices in BF ironmaking process. First, since the widely used least-square support regression (LS-SVR) algorithm cannot cope with the multi-output problem directly, this paper uses multi-task transfer learning technology to construct a novel multi-output LS-SVR (M-LS-SVR) for multivariable nonlinear systems. Then, this M-LS-SVR is adopted to identify the inverse system model of the controlled BF ironmaking process with the help of the presented modeling performance comprehensive evaluation and NSGA II based multi-objective parameter optimization. In order to better perform the control of the MIQ indices, the identified inverse system model is used to compensate the controlled nonlinear BF system to a compounded pseudo linear system with linear transitive relation. Such an inverse system based data-driven predictive control can effectively improve the control performance of the conventional nonlinear predictive control. Data experiments using actual industrial data from a large BF show that the proposed methods are effective, advanced and practical, and provide a solution to the operational control and optimization of the BF ironmaking process.

Matrix completion is widely used in many practical applications such as computer vision and data mining. In this paper, we consider the following two issues arising in real scenarios. First, the collected datasets are damaged by sparse noise and column outliers simultaneously; second, the datasets are not static in nature due to the existing of out-of-sample. Both of them have been ignored by most existing methods. In contrast with the traditional matrix completion algorithms which aim at recovering the entire matrix directly, we in this paper aim to first learn a low dimensional subspace by recovering a subset of collected samples, and then utilize it to estimate the missing values of residual data. There are two important advantages about this transformation. First, weakening the deviation caused by column outliers. Second, providing a direction for efficiently solving the out-of-sample problem. Particularly, to further improve the robustness of proposed method to sparse noise, we present a novel robust matrix completion model and a robust vector completion model, and both of them are based on non-convex lp-norm (0 In experiments, the proposed method and other state-of-the-art algorithms will be used to cope with two problems: matrix completion and vector completion. Numerical results on real datasets and artificial datasets demonstrate that our method can provide a significant performance advantage over alternative methods.

The automatic segmentation of variations in fundus images is being increasingly developed to establish retinal health, and to diagnose and monitor retinal abnormality. Fundus images can be acquired by low-cost and easy access diagnostic systems also in resource limited countries. In this paper, we focus on the segmentation of pigment signs in retinal fundus images, which is an important step in the diagnosis and monitoring of Retinitis Pigmentosa. At present, most of the existing approaches adopted for this purpose are based on the analysis of Optical Coherence Tomography (OCT) data, with only a few algorithms working on fundus images. The contribution of this paper is twofold. First, we propose a supervised segmentation technique for pigment sign detection in fundus images, which exploits the ensemble classifiers for both bagged and boosted decision trees, namely Random Forests and AdaBoost.M1. The ensemble classifiers are trained on feature vectors encoding the information of a set of regions obtained by partitioning a pre-processed image, so that each region is labeled as either normal fundus or pigment sign. Secondly, we have collected a large dataset of retinal images, in which the pigment signs have been segmented manually and we are making this publicly available to the research community. The performance of the classifiers has been evaluated on the proposed dataset and can be considered as a baseline for comparison. Our results demonstrate the high effectiveness of a machine learning approach as a suitable tool for automated Retinitis Pigmentosa analysis. The automatic segmentation of variations in fundus images is being increasingly developed to establish retinal health, and to diagnose and monitor retinal abnormality. Fundus images can be acquired by low-cost and easy access diagnostic systems also in resource limited countries. In this paper, we focus on the segmentation of pigment signs in retinal fundus images, which is an important step in the diagnosis and monitoring of Retinitis Pigmentosa. At present, most of the existing approaches adopted for this purpose are based on the analysis of Optical Coherence Tomography (OCT) data, with only a few algorithms working on fundus images. The contribution of this paper is twofold. First, we propose a supervised segmentation technique for pigment sign detection in fundus images, which exploits the ensemble classifiers for both bagged and boosted decision trees, namely Random Forests and AdaBoost.M1. The ensemble classifiers are trained on feature vectors encoding the information of a set of regions obtained by partitioning a pre-processed image, so that each region is labeled as either normal fundus or pigment sign. Secondly, we have collected a large dataset of retinal images, in which the pigment signs have been segmented manually and we are making this publicly available to the research community. The performance of the classifiers has been evaluated on the proposed dataset and can be considered as a baseline for comparison. Our results demonstrate the high effectiveness of a machine learning approach as a suitable tool for automated Retinitis Pigmentosa analysis.

This paper addresses the consensus control problem for a class of multi-agent systems with distributed parameter models, and all the agents in the considered systems are governed by the one-dimensional parabolic equations or the one-dimensional second-order hyperbolic equations. The main contribution of this paper is to apply Lyapunov functional approach to the multi-agent systems with distributed parameter models, and solve the consensus control problem of such multi-agent systems on an appropriate Sobolev space. The distributed consensus-based feedback control protocols are obtained based on the virtual leader approach, and when the feedback control laws are applied to the systems, consensus on L2(0, 1) space (corresponding to the parabolic equations) or on W1, 2(0, 1) × L2(0, 1) space (corresponding to the hyperbolic equations) is achieved for all the directed communication graphs with spanning trees. Simulation examples illustrate the effectiveness of the proposed method.

This paper presents novel iterative learning methods along with the neural network implementation for multi-player zero-sum games. Solving zero-sum games depends on the solutions of Hamilton–Jacobi-Isaacs equations, which are nonlinear partial differential equations. These solutions are generally difficult or even impossible to be obtained analytically. To overcome this difficulty, iterative adaptive dynamic programming algorithms are utilized. In the related research works, three-network architecture, i.e., critic-actor-disturbance structure, is used to approximate the value function, control policies and disturbance policies. Different from the previous works, this paper employs single-network architecture, i.e., critic-only structure, to implement the proposed algorithms, which reduces the computation burden and the complexity of design procedure. Finally, two simulation examples are provided to illustrate the effectiveness of our proposed methods.

This paper introduces an innovative interpretation of spiking signal processing (SSP) and proposes applications in control system. Taking the firing rate as the coding principle employed by biological neurons, we have transformed continuous signals in geometric spiking series evolving over time. New fundamental SSP equations are presented and compared to classical digital signal processing (DSP). Application to the linear system analysis and control drive using spiking transformation is finally presented. This paper launches the theoretical SSP basis to investigate more deeply the geometric neural networks with learning ability.

Activation function is crucial to the recent successes of deep neural networks. In this paper, we first propose a new activation function, Multiple Parametric Exponential Linear Units (MPELU), aiming to generalize and unify the rectified and exponential linear units. As the generalized form, MPELU shares the advantages of Parametric Rectified Linear Unit (PReLU) and Exponential Linear Unit (ELU), leading to better classification performance and convergence property. In addition, weight initialization is very important to train very deep networks. The existing methods laid a solid foundation for networks using rectified linear units but not for exponential linear units. This paper complements the current theory and extends it to the wider range. Specifically, we put forward a way of initialization, enabling training of very deep networks using exponential linear units. Experiments demonstrate that the proposed initialization not only helps the training process but leads to better generalization performance. Finally, utilizing the proposed activation function and initialization, we present a deep MPELU residual architecture that achieves state-of-the-art performance on the CIFAR-10/100 datasets. The code is available at https://github.com/Coldmooon/Code-for-MPELU .

This paper proposes a methodology for developing an adaptive neural network controller for a simulated helicopter system. Since an indirect adaptive neural network framework is chosen, the controller comprises three interconnected neural networks called the observer, actor and critic. The actor and critic networks rely on the observer network responsible for state estimation. The main contribution of this paper is the development of an observer that has fast convergence capabilities in order to be used in a completely on-line stability control scheme. This improved convergence is obtained by uniquely modifying the observer network structure and update law. The observer parameters are also optimised by means of a genetic algorithm (GA) for improved performance. The developed observer is firstly evaluated on a first principle linear model and then on actual test flight data of an attack helicopter. The results indicate excellent performance in terms of the state estimation capability of the observer. Lyapunov’s direct method is used to derive update laws for both the critic and actor networks and the control parameters of these networks are also optimised by means of a multi-objective GA. Actual data from a wind-tunnel test set-up were used for controller evaluation purposes.

Pedestrian detection and tracking have become an important field in the computer vision research area. This growing interest, started in the last decades, might be explained by the multitude of potential applications that could use the results of this research field, e.g. robotics, entertainment, surveillance, care for the elderly and disabled, and content-based indexing. In this survey paper, vision-based pedestrian detection systems are analysed based on their field of application, acquisition technology, computer vision techniques and classification strategies. Three main application fields have been individuated and discussed: video surveillance, human-machine interaction and analysis. Due to the large variety of acquisition technologies, this paper discusses both the differences between 2D and 3D vision systems, and indoor and outdoor systems. The authors reserved a dedicated section for the analysis of the Deep Learning methodologies, including the Convolutional Neural Networks in pedestrian detection and tracking, considering their recent exploding adoption for such a kind systems. Finally, focusing on the classification point of view, different Machine Learning techniques have been analysed, basing the discussion on the classification performances on different benchmark datasets. The reported results highlight the importance of testing pedestrian detection systems on different datasets to evaluate the robustness of the computed groups of features used as input to classifiers.

In this paper, we present the analog circuits to solve a class of variational inequality problems (VIPs) based on the projection neural network (PNN) and inertial projection neural network (IPNN) algorithms. The proposed circuits are normative and only require basic circuit elements. The optimal solutions of VIPs are equivalent to the stable output voltages of the associated circuits. This paper also shows how to design analog circuits with projection operators (box constrains set and sphere constrains set) on the basis of PNN and IPNN algorithms. As a result, a class of variational inequality problems can be solved by proposed circuit frameworks. The effectiveness and superiority (with less computing time) of the proposed analog circuits are expound by simulating on three examples.

The paper considers a class of nonsmooth neural networks where hard-limiter saturation nonlinearities are used to constrain solutions of a linear system with concentrated and distributed delays to evolve within a closed hypercube of R n . Such networks are termed delayed linear systems in saturated mode (D-LSSMs) and they are a generalization to the delayed case of a relevant class of neural networks previously introduced in the literature. The paper gives a rigorous foundation to the D-LSSM model and then it provides a fundamental result on convergence of solutions toward equilibrium points in the case where there are nonsymmetric cooperative (nonnegative) interconnections between neurons. The result ensures convergence for any finite value of the maximum delay and is physically robust with respect to perturbations of the interconnections. More importantly, it encompasses situations where there exist multiple stable equilibria, thus guaranteeing multistability of cooperative D-LSSMs. From an application viewpoint the delays in combination with the property of multistability make D-LSSMs potentially useful in the fields of associative memories, motion detection and processing of temporal patterns.

Data-driven algorithms for bearing fault diagnosis have achieved much success. However, it is difficult and even impossible to collect enough data containing real bearing damages to train the classifiers, which hinders the application of these methods in industrial environments. One feasible way to address the problem is training the classifiers with data generated from artificial bearing damages instead of real ones. In this way, the problem changes to how to extract common features shared by both kinds of data because the differences between the artificial one and the natural one always baffle the learning machine. In this paper, a novel model, deep inception net with atrous convolution (ACDIN), is proposed to cope with the problem. The contribution of this paper is threefold. First and foremost, ACDIN improves the accuracy from 75% (best results of conventional data-driven methods) to 95% on diagnosing the real bearing faults when trained with only the data generated from artificial bearing damages. Second, ACDIN takes raw temporal signals as inputs, which means that it is pre-processing free. Last, feature visualization is used to analyze the mechanism behind the high performance of the proposed model.

The purpose of this paper is to introduce a new approach for the free-hand sketch representation in the sketch based image retrieval (SBIR), where the sketches are treated as the queries to search for the natural photos in the natural image dataset. This task is known as an extremely challenging work for 3 main reasons: (i) sketches show a highly abstract visual appearance versus natural photos, fewer context can be extracted as descriptors using the existing methods, (ii) for the same object, different people provide widely different sketches, making sketch-photo matching harder, (iii) mapping the sketches and photos into a common domain is also a challenging task. In this paper, we address the cross-domain question using a strategy of mapping sketches and natural photos in multiple layers. For the first time, we introduce a multi-layer deep CNNs framework to train the multi-layer representation of free hand sketches and natural photos. We use Flickr15k dataset as benchmark for the retrieval and show that our learned representation significantly outperformances both hand-crafted features as well as deep features trained by sketches or photos.

This paper presents a novel hyperspectral image (HSI) classification framework by exploiting multi-scale spectral-spatial features via hierarchical recurrent neural networks. The neighborhood information plays an important role in the image classification process. Convolutional neural networks (CNNs) have been shown to be effective in learning the local features of HSI. However, CNNs do not consider the spatial dependency of non-adjacent image patches. Recurrent neural networks (RNNs) can effectively establish the relationship of non-adjacent image patches, but it can only be applied to single-dimensional (1D) sequence. In this paper, we propose multi-scale hierarchical recurrent neural networks (MHRNNs) to learn the spatial dependency of non-adjacent image patches in the two-dimension (2D) spatial domain. First, to better represent the objects with different scales, we generate multi-scale 3D image patches of central pixel and surrounding pixels. Then, 3D CNNs extract the local spectral-spatial feature from each 3D image patch, respectively. Finally, multi-scale 1D sequences in eight directions are constructed on the 3D local feature domain, and MHRNNs are proposed to capture the spatial dependency of local spectral-spatial features at different scales. The proposed method not only considers the local spectral-spatial features of the HSI, but also captures the spatial dependency of non-adjacent image patches at different scales. Experiments are performed on three real HSI datasets. The results demonstrate the superiority of the proposed method over several state-of-the-art methods in both visual appearance and classification accuracy.

The typical model selection strategy applied in most Extreme Learning Machine (ELM) papers depends on a k-fold cross-validation and a grid search to select the best pair {L, C} of two adjustable hyper-parameters, namely the number L of hidden ELM nodes and the regularisation parameter, C, that minimizes the validation error. However, by testing only 30 values for L and 30 values for C via 10-fold cross-validation, the learning phase must build 9000 different ELM models, each with a different pair {L, C}. Since these models are not independent from one another, the essence of managing and drastically reducing the computational cost of the ELM model selection relies on matrix decompositions that avoid direct matrix inversion and allow producing reusable matrices during the cross-validations. Still, one can find many matrix decompositions and cross-validation versions that result in several combinations. In this paper, we identify these combinations and analyse them theoretically and experimentally to discover which is the fastest. We compare Singular Value Decomposition (SVD), Eigenvalue Decomposition (EVD), Cholesky decomposition, and QR decomposition, which produce re-usable matrices (orthogonal, Eigen, singular, and upper triangular). These decompositions can be combined with different cross-validation approaches, and we present a direct and thorough comparison of many k-fold cross-validation versions as well as leave-one-out cross-validation. By analysing the computational cost, we demonstrate theoretically and experimentally that while the type of matrix decomposition plays one important role, another equally important role is played by the version of cross-validation. Finally, a scalable and computationally-effective algorithm is presented that significantly reduces computational time.

With the increasing demand of security and safety assurance for public, anomaly detection has gained a greater focus in the field of intelligent video surveillance analysis. In this paper, a novel method is proposed to address the issue in anomaly detection. It is based on Nearest Neighbor (NN) search with the Locality-Sensitive B-tree (LSB-tree), which helps to find the approximate NNs among the normal feature samples for each test sample. To better analyze the pedestrian behavior, not only the commonly used motion-appearance feature is applied in the method, but also a novel feature is proposed to describe the dynamic changes of the behavior. Compared to the relative works, the main novelties of this paper mainly includes: (1) the method of LSB-tree, which enables fast high-dimensional NN search, is applied in this paper to evaluate the similarity between the test samples and normal feature samples; (2) in order to analyze the dynamic motion and appearance, the Dynamics of Pedestrian Behavior (DoPB) feature on Riemannian manifolds is applied as the individual descriptor, which helps to detect the drastic behaviors and abnormal translation motions; (3) a new evaluation method is developed to generate the anomaly map and determine the anomaly. Experimental results and the comparisons with state-of-the-art methods demonstrate that the proposed method is effective in anomaly detection and localization, and is applicable in various scenes.

In this paper, the problem of feedback controller designed to achieve adaptive exponential stabilization is investigated for stochastic neural networks of neutral type with Markovian switching parameters and the noise characterised by Levy process. Based on Lyapunov functional theory and the generalized Itos formula for neutral-type systems, the goal of this paper is to derive some criteria to ensure adaptive exponential stabilization for the stochastic neutral-type neural networks. Moreover, the update law of the control gain and the dynamic variation of the parameters of the system are provided. Finally, the theoretical analysis and potential of the stabilization criteria proposed in this paper are verified by a numerical example.

Based on the fact that actual cerebral cortex has different structure, a new heterogeneous simplified pulse coupled neural network (HSPCNN) model is proposed in this paper for image segmentation. HSPCNN is constructed with several simplified pulse coupled neural network (SPCNN) models, which have different parameters corresponding to different neurons. An image is segmented by HSPCNN into several regions according to their gray levels. Moreover, the parameter of HSPCNN is set automatically in this paper, the experimental segmentation results of the gray natural images from the Berkeley Segmentation Dataset (BSD 300) show the validity and efficiency of the proposed segmentation method. Finally, an evaluation index is proposed to measure the segmentation result.

A small dataset often makes it difficult to build a reliable learning model, and thus some researchers have proposed virtual sample generation (VSG) methods to add artificial samples into small datasets to extend the data size. However, for some datasets the assumption of the distribution of data in the VSG methods may be vague, and when data only has a few attributes, such approaches may not work effectively. Other researchers thus proposed attribute extension methods to generate attributes to convert data into a higher dimensional space. Unfortunately, the resulting dataset may become a sparse dataset with many null or zero values in extended attributes, and then a large quantity of such attributes will reduce the representativeness of instances for the learning model. Therefore, based on fuzzy theories, this paper proposes a novel sample attribute extending (SEA) method to extend a suitable quantity of attributes to improve small dataset learning. In order to verify the validity of the SEA method, using SVR and BPNN, this paper adopts two real cases and two public datasets to conduct the learning of the predictive model, and uses the paired t-test to statistically examine the significance of improvement. The experimental results show that the proposed SEA method can effectively improve the learning accuracy of small datasets.

Neural Networks and Support Vector Machines (SVMs) are two of the most popular and efficient supervised classification models. However, in the context of large datasets many complexity issues arise due to high memory requirements and high computational cost. In the context of the application of Data Mining algorithms, data reduction techniques attempt to reduce the size of training datasets in terms of the number of instances by selecting some of the existing instances or by generating new training instances. The idea is to speed up the application of the data mining algorithm with minimum or no sacrifice in performance. Data reduction techniques have been extensively used in the context of k-Nearest Neighbor classification, a lazy classifier that works by directly using a training dataset rather than building a model. This paper explores the application of data reduction techniques as a preprocessing step before the training step of Neural Networks and SVMs. Furthermore, the paper proposes a new data reduction technique that is based on k-median clustering algorithm. Our experimental results illustrate that, in the case of SVMs, data reduction techniques can effectively reduce the dataset size incurring small performance degradation. In the case of Neural Networks, the performance loss is somewhat greater, for the same data reduction rate, but both SVM and Neural Network models outperform the k-NN approach that is typically used in Data Mining applications.

This paper investigates the optimal control of uncertain nonlinear systems with external disturbance via adaptive dynamic programming, where two kinds of controllers are introduced and event-driven strategy is used to update the controllers as well as the network weights. One controller is constrained-input, and another is used to offset the external disturbance. Three kinds of neural networks (identifier network, critic network and actor network) are used to approximate the uncertain dynamics, the optimal value, control inputs and external disturbance, respectively. To the best knowledge of the authors, the advantage of this paper is that almost all the existing systems about optimal control via adaptive dynamic programming can be considered as special cases of this system, where the external disturbance is considered. By designing appropriate parameters, the uncertain nonlinear system is asymptotically stable with the event-driven controller. And, under the given learning rates of neural networks, the system state and the estimation error of neural networks are proved to be uniform ultimate boundedness (UUB) via adaptive dynamic programming. Finally, numerical simulation results are presented to verify the theoretical analysis.

This paper investigates an adaptive neural network prescribed performance synchronization scheme for unknown complex dynamic networks with different dimensions. Based on predefined performance bounded and Lyapunov stability theory, adaptive neural networks controllers are designed to ensure that synchronization errors remain in a neighborhood of origin with the prescribed bounds. In addition, the paper analyses in detail that the synchronization behaviors between drive network selected as the three-dimension chaotic system and response network selected as the four-dimension hyperchaotic chaotic system. The numerical simulation results are presented to show the validity of the proposed approach.

In this paper, a fully distributed output regulation problem is investigated for a linear multi-agent system with both cooperative and competitive interactions. The interaction network associated with the multi-agent system is conveniently modeled by an undirected signed graph and called coopetition network for simplicity. In most literatures on cooperative output regulation problems, it is commonly assumed that the state matrix of the exogenous system is available to all agents. In this paper, a distributed adaptive observer is developed for each agent to estimate the state and the state matrix of the exogenous system to relax the constraint. Furthermore, a reduced-order observer based controller is proposed for each agent by using an internal model approach to guarantee bipartite consensus. Finally, simulation results are provided to demonstrate the proposed control strategy.

Computer simulation, which has been proved to be an effective approach to problem solving, is nowadays widely used in modern science. However, it requires a lot of computing resources, which are difficult for general users to acquire. In this paper, we design a Web based system to implement on-line simulation system for ordinary users. As a useful example, the simulation of one type of collision detection model is presented in this paper. Moreover, the software application of simulation is offered as a service on Web. Meanwhile, the incorporation of general vector machine (GVM, a type of neural network) to intelligently predict the relationship between simulation parameters and computation resources is presented, which could further provide more information for system monitoring and scheduling. The system has demonstrated efficiency and intuitiveness for users of this type of applications.

An important feature characteristic of the data streams in many of today's big data applications is the intrinsic uncertainty, which could happen for both item occurrence and attribute value. While this has already posed great challenges for fundamental data mining tasks such as classification, things are made even more complicated by the fact that completely-labeled examples are usually unavailable in such settings, leaving researchers the only option to learn classifiers on partially-labeled examples on uncertain data streams. Furthermore, there will be concept drift on evolving data streams. To address these challenges, this paper therefore focuses on the study of learning from positive and unlabeled examples (PU) on uncertain data streams. To the best of our knowledge, this paper is the first work to address the uncertainty issue in both item occurrence (occurrence level) and attribute value (attribute level) for the problem of PU learning over streaming data. Firstly, we propose an algorithm to classify positive and unlabeled examples with both uncertainties (PUU). The algorithm extracts reliable positive and negative examples by clustering-based method, and then trains the classifier with Weighted Extreme Learning Machine (Weighted ELM). Secondly, we propose an algorithm of PU learning over uncertain data streams (PUUS). It adopts ensemble model and trains base classification by PUU. In order to detect concept drift, PUUS uses cluster set similarities between the current data block and history data block. We propose different update strategies for different concept drift to adapt to evolving uncertain data streams. Experimental results show that PUUS can effectively classify uncertain data streams with just positive and unlabeled data, while achieving in the meantime good performance in detecting and handling concept drifts.

In this paper, we propose an innovative classification method which combines texture features of images filtered by Gaussian derivative models with extreme learning machine (ELM). In the texture image classification, feature extraction is a very crucial step. Thusly, we use linear filters consisting of two Gaussian derivative models, difference of Gaussian (DOG) and difference of offset Gaussian (DOOG), to detect texture information of images. Besides, ensemble extreme learning machine (E2LM) is proposed to reduce the randomness of original ELM and used as the classifier in this paper. We evaluate the performance of both the texture features and the classifier E2LM by using three datasets: Brodatz album, VisTex database and Berkeley image segmentation database. Experimental results indicate that Gaussian derivative models are superior to Gabor filters, and E2LM outperforms the support vector machine (SVM) and ELM in classification accuracy.

This paper investigates the finite-time robust synchronization problem of mixed time-delayed neural networks (DNNs) with discontinuous activations and uncertain external perturbations. Two classes of simple discontinuous controllers (i.e., switching state-feedback controller and switching adaptive controller) are designed such that the response system can be finite-time robustly synchronized with a drive system. Without using the well-known finite-time stability theorem given by Forti in 2006, the finite-time robust synchronization control of discontinuous DNNs can be realized via generalized Lyapunov–Krasovskii functionals approach and novel analysis techniques. Moreover, the upper bounds of the settling time for synchronization are estimated. The main tools of this paper involve functional differential inclusion theory in the sense of Filippov solutions and non-smooth analysis. Finally, two numerical examples are provided to illustrate the effectiveness of the designed control method and the theoretical results.

In this paper, the problem of H∞ consensus for linear heterogeneous multi-agent systems (HMASs) with state and output feedback control is investigated. By using the linear matrix inequality (LMI) theory, some basic properties of symmetric positive definite matrix and a series of analysis, the novel consensus conditions of using state and output feedback control and the corresponding algorithm are provided. The key idea of this paper is to convert one kind of nonlinear matrix inequality condition into a series of novel conditions that can be solved by using LMI toolbox and the most important characteristic of this paper is that the necessary and sufficient conditions of H∞ consensus are provided while only sufficient conditions were given in some existing results. First, an equivalent condition for a special kind of nonlinear matrix inequality is given. Second, by using this equivalent condition, the necessary and sufficient conditions of H∞ consensus for linear HMASs under state and output feedback control approaches are provided, respectively. Third, according to iterative algorithm and the property of trace of matrix, a feasible algorithm for calculating the controllers is proposed. Fourth, the obtained results are extended to solve the H∞ control problem for general linear system with some additional conditions. At last, two numerical examples are given to support our results.

In this paper, a hierarchical method for human body segmentation in static images is proposed. Based on the face detection, a torso model is developed to locate the torso to provide the foreground seeds for torso segmentation with graph cuts. Similarly, an upper leg model is designed for lower body segmentation based on the segmented torso. Besides, a general method which combines shape information into graph cuts is presented, which can make the segmentation more accurate. The main contributions of this paper include two models and a new method of integrating shape information into graph cuts. Experiments on our collected real world images show that our method can accurately recover human body from static images with a variety of individuals, poses, backgrounds and clothing.

The tumour suppressor gene p53 plays a key role in cell response to DNA damage, and the p53Mdm2 regulation relationship is crucial for the expression of p53. In this paper, based on gene expression time-series data of human leukaemia cells after exposure to ionising radiation, with ionising radiation as input, a nonlinear continuous time-delay dynamic stochastic mathematical model of a p53Mdm2 network is established using a continuousdiscrete extended Kalman filter algorithm. The accuracy of the established model is then validated. Numerical simulation is used to simulate the dynamic regulation in p53Mdm2 networks for low, medium, and high ionising radiation doses. The results show that the proposed algorithm is convergent and that the error rate of the model is only 1.19%. In addition, the model can simulate accurately the response of the p53Mdm2 network to different doses of ionising radiation. The methods proposed in this paper can supply the foundation for research on the dynamics of the p53Mdm2 gene regulation relationship and play a guiding role in research on the p53Mdm2 response process after DNA damage under different doses of ionising radiation.

Facial expression recognition is an important research issue in the pattern recognition field. In this paper, we intend to present a novel framework for facial expression recognition to automatically distinguish the expressions with high accuracy. Especially, a high-dimensional feature composed by the combination of the facial geometric and appearance features is introduced to the facial expression recognition due to its containing the accurate and comprehensive information of emotions. Furthermore, the deep sparse autoencoders (DSAE) are established to recognize the facial expressions with high accuracy by learning robust and discriminative features from the data. The experiment results indicate that the presented framework can achieve a high recognition accuracy of 95.79% on the extended Cohn–Kanade (CK+) database for seven facial expressions, which outperforms the other three state-of-the-art methods by as much as 3.17%, 4.09% and 7.41%, respectively. In particular, the presented approach is also applied to recognize eight facial expressions (including the neutral) and it provides a satisfactory recognition accuracy, which successfully demonstrates the feasibility and effectiveness of the approach in this paper.

Aiming at the complexity and particularity of urban environment, a solar-powered UAV (SUAV) path planning framework is proposed in this paper. The framework can be decomposed into three aspects to resolve. First, to make SUAV avoid the building obstacles, a nature-inspired path planning method called Interfered Fluid Dynamical System (IFDS) is introduced. Aiming at the defect that the traditional IFDS is not suitable for SUAV energy optimization calculation, the dynamic constraints and model are introduced to IFDS. The modified IFDS, called Restrained IFDS (RIFDS), is proposed. Second, to resolve the path planning issue efficiently, a novel intelligent optimization algorithm called Whale Optimization Algorithm (WOA) is selected as the basic framework solver. To further overcome the drawback of local minima, adaptive chaos-Gaussian switching solving strategy and coordinated decision-making mechanism are introduced to the basic WOA. The modified algorithm, called Improved WOA (IWOA), is proposed. Third, to solve the accurate modeling problem of solar energy in urban environment, two measures are adopted: (1) A practical judgment method for sunlight occlusions is proposed; (2) Aiming at some unreasonable aspects in the solar energy production model, the received solar energy is modified and recalculated by ASHRAE Clear Sky Model and the solar irradiance calculation principle for slant surfaces in this paper. Finally, the effectiveness of the proposed framework is tested by the simulations.

This paper addresses the problem of global exponential stability for a class of memristive Cohen–Grossberg with mixed time delays and impulsive time window. Based on the memristor theory, impulse control theory, and mathematical induction method, stability criteria for the considered memristive Cohen–Grossberg neural networks are derived by employing appropriate Lyapunov functions. Compared with existing impulse control schemes, the impulse instants are extended to some bounded time intervals, we will show that the neural networks can still be stable under reasonable assumptions. Furthermore, the conditions obtained in this paper are easy to be checked, and they can be applied to improve previous results concerning stability for memristive neural networks. Finally, a numerical example is given to illustrate the effectiveness of the theoretical results.

Histogram based image enhancement techniques are widely used for performing contrast enhancement in images. However, most histogram based image enhancement methods have insufficient capability to freely tune the brightness and contrast of enhanced image. In this paper, two novel histogram based image enhancement algorithms are proposed. The proposed algorithms provide the way to control the brightness and contrast of enhanced image by adjusting two parameters. The principles for parameter selection are also discussed in this paper. Experimental results demonstrate a better performance of the proposed methods in both perceptual quality and image quality assessment metrics than the existing histogram based methods.

This paper proposes the design and implementation of a model-free tire slip control for a fast and highly nonlinear Anti-lock Braking System (ABS). A reinforcement Q-learning optimal control approach is inserted in a batch neural fitted scheme using two neural networks to approximate the value function and the controller, respectively. The transition samples required for learning high performance control can be collected by interacting with the process either by online exploiting the current iteration controller (or policy) under an e-greedy exploration strategy, or by using data collected under any other controller that is capable of ensuring efficient exploration of the action-state space. Both approaches are highlighted in the paper. Fortunately, the ABS process fits this type of learning-by-interaction because it does not need an initial stabilizing controller. The validation case studies conducted on a real laboratory setup reveal that high control system performance can be achieved using the proposed approaches. Insightful comments on the observed control behavior are offered along with performance comparisons with several types of model-based and model-free controllers including relay, model-based optimal PI, an original model-free neural network state-feedback VRFT controller and a model-free neural network adaptive actor-critic one. With the ability to improve control performance starting from different supervisory controllers or to learn high performance controllers from scratch, the proposed Q-learning optimal control approach proves its performance in a wide operating range and is therefore recommended to its industrial application on ABS.