Search

This paper investigates the distributed output-feedback tracking problem for a class of nonlinear multi-agent systems (MASs) with polynomial growth in unmeasurable states. Essentially different from the case with linear growth, it drives us to develop a more powerful control policy to cope with the polynomially growing nonlinearities. To ensure that the error between the follower’s output and the leader’s output is arbitrarily small while all the states of the closed-loop system are globally bounded, a nominal system is considered first. For this, we establish a distributed state-feedback virtual controller firstly to obtain the controller structure for the considered control problem, and then, by the distributed high-gain homogeneous domination method, a homogeneous output-feedback controller is obtained. On the basis of the available tracking controller, we turn to the control problem itself and provide the performance analysis. To be emphasized, the proposed distributed state-feedback virtual controller enables us to achieve rather general results on global output-feedback stabilization of nonlinear multi-agent systems with polynomial growth. Finally, the effectiveness of the design scheme is illustrated by an example.

Self-supervised contrastive learning can help alleviating the meet of large numbers of annotated samples and learning high-level representations from unlabeled data. However, the high diversities in ground objects make it difficult to learn the features at more robust and refined level in synthetic aperture radar (SAR) image analysis. To alleviate this issue, we propose a self-supervised weighted contrastive learning method with context-augmented transformer for change detection in multiresolution SAR images. First, a weighted contrastive learning framework is built by introducing a weighted contrastive loss, which can reduce the influence of changed pixels in the process of self-supervised feature learning and align feature representations of image pairs. Then, to model complex and rich context information, a context-augmented swin transformer is proposed to aggregate contextual information and compute hierarchical representations, which are beneficial for dense prediction. Specially, global channel-wise aggregation module and multiscale fusion structure are designed to enhance global features and capture fine-scale features, respectively. Thus, rich local, global and multiscale context information can be modeled jointly to achieve fine and robust feature expression. Compared with other network, our network gives full play to the advantages of CL and transformer to extract representations with rich context information in unsupervised scenes, with good generalization. Experiments on real SAR images with different resolutions demonstrate the effectiveness and superiority of the proposed method.

Under a directed leader-follower network structure, this manuscript addresses the cooperative control issue for a class of stochastic upper-triangular nonlinear multi-agent systems (MASs) with unknown growth rates. Initially, by applying time-varying coordinates, the system is converted into an equivalent form with time-dependent gain factors. Subsequently, a time-varying state feedback controller is developed, demonstrating that followers can achieve mean-square bounded output tracking of the leader, with maintaining all the closed-loop system bounded in probability. Lastly, the proposed approach’s effectiveness is confirmed through simulation experiments.

The advent of the sixth generation of mobile communications (6G) ushers in an era of heightened demand for advanced network intelligence to tackle the challenges of an expanding network landscape and increasing service demands. Deep Learning (DL), as a crucial technique for instilling intelligence into 6G, has demonstrated powerful and promising development. This paper provides a comprehensive overview of the pivotal role of DL in 6G, exploring the myriad opportunities and challenges that arise. Firstly, we present a detailed vision for DL in 6G, emphasizing areas such as adaptive resource allocation, intelligent network management, robust signal processing, ubiquitous edge intelligence, and endogenous security. Secondly, this paper reviews how DL models leverage their unique learning capabilities to solve complex service demands in 6G. The models discussed include Convolutional Neural Networks (CNN), Generative Adversarial Networks (GAN), Graph Neural Networks (GNN), Deep Reinforcement Learning (DRL), Transformer, Federated Learning (FL), and Meta Learning. Additionally, we examine the specific challenges each DL model faces within the 6G context. Moreover, we delve into the rapidly evolving field of Artificial Intelligence Generated Content (AIGC), examining its development and impact within the 6G framework. Finally, this paper culminates in a detailed discussion of ten critical open problems in integrating DL with 6G, setting the stage for future research and development in this field.

The carrier multiplication phenomenon involves hot carriers, which gain kinetic energy while accelerating to equilibrium with the established avalanching electric fields, and is typically explained via the local avalanche model. This work presents two vertical Ge-on-Si avalanche photodetectors fabricated in a separate absorption, charge, and multiplication configuration. Uniformity in materials, doping densities, and device dimensions is maintained, except for the multiplication width, which is used as a control parameter to manipulate avalanching fields under identical electric biasing and illumination schemes. Nonlocal carrier multiplication model is implemented during analysis of the extracted current-voltage signatures under small and large reverse biasing arrangements. For such an APD characterized by thinner multiplication region $\left ({{W_{m}=0.1 ~\mu m}}\right)$ , reduced linear and Geiger-mode multiplication regimes are perceived to be at play, outperforming the device having thicker multiplication region in almost all related figures of merit, e.g., responsivity $(22.58 \mathrm {A/W})$ , photo-to-dark current ratio $(\sim {10}^{5})$ , normalized photo-to-dark current ratio $(2.5\times {10}^{9} \mathrm {W}^{-1})$ , specific detectivity $(7.45\times {10}^{12}\mathrm {Jones})$ , and noise equivalent power $(\sim 2.42\times {10}^{-15} \mathrm {W/}\sqrt {\mathrm {Hz}})$ . The enhanced performance characteristics are due to excessively strong avalanching fields, reduced thermal charge density, and negligible dead space compared to its counterpart characterized by thicker multiplication width.

With the maturity of satellite imaging technology, satellite video has attracted more and more attention because of high spatial resolution and temporal resolution. Thus, object tracking as the main application task of satellite videos also becomes a popular research topic. Compared to natural videos, the difficulties of satellite videos object tracking mainly are caused by feature deficiency of small objectives, surroundings, occlusion (OCC), etc. In this article, based on a dual correlation filter (DCF) tracking framework, a new object tracking method for satellite videos is proposed to deal with the above-mentioned problems. For feature deficiency problem, the proposed super-resolution feature enhancement module could improve the feature discriminative ability of objects utilizing the prior information learned by a super-resolution network. For the OCC problem, a multilayer perceptron motion prediction network is designed to predict the position of objects when OCC occurs, which uses an online training strategy. Besides, KNN background subtraction also is introduced to reduce the interference of surroundings. Finally, these above-mentioned processes are combined together appropriately on the DCF tracking framework for better tracking results. Aiming to verify the performance of the proposed method, abundant experiments have been conducted on satellite video datasets. The experimental results show that the design of the proposed tracking method is effective and it also has a conspicuous advantage compared with some state-of-the-art methods.

This article presents the scientific outcomes of the 2023 Data Fusion Contest (DFC23) organized by the Image Analysis and Data Fusion Technical Committee of the IEEE Geoscience and Remote Sensing Society. The contest consists of two tracks investigating the fusion of optical and synthetic aperture radar data for: 1) fine-grained roof type classification and 2) height estimation. During the development phase, 1000 people registered for the contest, while at the end 55 and 35 teams competed during the test phase in the two tracks, respectively. This article presents the methods and results obtained by the first and second-ranked teams of each track. In Track 1, both winning teams leveraged pretraining, modern network architectures, model ensembles, and measures to cope with the imbalanced class distribution. The solutions to Track 2 are more diverse and are characterized by modern multitask learning approaches. The data of this contest is openly available to the community for further research, development, and refinement of machine learning methods.

In recent years, constrained by the challenges associated with expensive data annotation and poor generalization ability in supervised models, domain adaptation has been proposed to effectively mitigate domain gaps, which has gained significant attention in the field of remote sensing. However, most existing methods do not consider the inherent characteristics of remote sensing images when formulating adaptive strategies. In addition, there has been limited research on addressing class-imbalanced data situations, leading to undesirable performance in domain adaptation tasks involving long-tailed datasets. To overcome the aforementioned limitations, based on the analysis of intraclass diversity and intradomain style differences of remote sensing images, we propose a novel prototype contrastive learning framework called general-specific prototype contrastive learning (GSPCL). Due to the unreliability of clustering samples in conventional prototype-based clustering methods, the bidirectional weighted prototype strategy is proposed to optimize this loophole. Consequently, more robust prototypes are constructed in both domains, serving as mediators to reduce domain discrepancies bidirectionally. Particularly, often overlooked in most methods, we incorporate low-confidence sample features into the contrastive learning process alongside these prototypes to further guide the model to address feature alignment and long-tail issues effectively. Finally, in order to verify the superiority of our proposed method, we adhere to two existing experimental settings and construct an extra optical remote sensing domain adaptation dataset with class-imbalanced scenarios. In the first two experimental settings, GSPCL outperforms the second-ranked approach by 5.0% and 4.0% in average accuracy. Furthermore, our approach exhibits highly competitive results in handling long-tailed data scenarios.

Complex-valued convolutional neural networks (CV-CNN) have recently gained recognition in feature representation learning. It implements the repeated application of the operations in convolution, local average pooling, and the absolute value of the resulting vectors. However, it is only conducted in the complex spatial domain, and lacks effective representation of directionality, singularity, and regularity in the complex spectral domain for anomaly detection of images. This is the key to feature learning representation of high-order singularity. To solve this problem, a complex-valued contourlet neural network (C$^{2}$N$^{2}$) is proposed in this article. It is novel in this sense that, different from the CV-CNN in the spatial domain, the spectral stream of C$^{2}$N$^{2}$ can enhance the multiresolution sparse representation of nonsubsampled contourlet (NSCT) with multiscales and multidirections for images. Furthermore, the spectral feature integration module is proposed to capture the statistical properties of the NSCT coefficients. It is shown that the proposed network can improve the distinguishability of feature learning and classification ability in theoretical analysis and experiments on three benchmark datasets (Flevoland, Xi'an, and Germany) compared with developed methods. Polarimetric synthetic aperture radar image classification is widely used in the fields of agriculture, forestry, and military. It must be emphasized that there is potential in effective feature learning representation and the generalization capability of C$^{2}$N$^{2}$ in deep learning, recognition, and interpretation.

The foundation model (FM) has garnered significant attention for its remarkable transfer performance in downstream tasks. Typically, it undergoes task-agnostic pretraining on a large dataset and can be efficiently adapted to various downstream applications through fine-tuning. While FMs have been extensively explored in language and other domains, their potential in remote sensing has also begun to attract scholarly interest. However, comprehensive investigations and performance comparisons of these models on remote sensing tasks are currently lacking. In this survey, we provide essential background knowledge by introducing key technologies and recent developments in FMs. Subsequently, we explore essential downstream applications in remote sensing, covering classification, localization, and understanding. Our analysis encompasses over 30 FMs in both natural and remote sensing fields, and we conduct extensive experiments on more than 10 datasets, evaluating global feature representation, local feature representation, and target localization. Through quantitative assessments, we highlight the distinctions among various FMs and confirm that pretrained large-scale natural FMs can also deliver outstanding performance in remote sensing tasks. After that, we systematically presented a brain-inspired framework for remote sensing foundation models (RSFMs). We delve into the brain-inspired characteristics in this framework, including structure, perception, learning, and cognition. To conclude, we summarize 12 open problems in RSFMs, providing potential research directions. Our survey offers valuable insights into the burgeoning field of RSFMs and aims to foster further advancements in this exciting area.

Automatic 2D vision-based defect detection on sealing nail (SealN) surfaces is challenging due to interference of complex backgrounds with non-homogeneous and low contrast between foreground and background. Inspired by an interesting observation that the albedo domain recovered by the uncalibrated photometric stereo (UPS) shows obvious differences and significant abruptness between defects’ and non-defects’ regions, we develop a novel semantic-guided variational model (SGVM) to conditional extract structural defects from albedo map. Specifically, SGVM utilizes one developed global regularized label indicator to semantically guide one local regularized relative Gaussian filter (RGF) for achieving large-scale structures (i.e., defects) preservation and small-scale textures (i.e., background) suppression. Furthermore, defects can be efficiently extracted by thresholding the structure map within the label indicator. Additionally, experimental results on numerous challenging defect images reveal that the proposed SGVM outperforms the existing advanced 2D methods in terms of defect extraction.

To improve the accuracy of graph neural network recommendation algorithms, research mainly integrates multi head attention mechanism and GRU, which is to construct a graph neural network recommendation model; Considering the long and short term preferences of users, a graph neural network algorithm integrating long and short term preferences is constructed. The research results indicated that when the embedding dimension was 64, the batch size of selected samples was 64, the learning rate was 0.0005, the vertical stacking layer of GRU was 2, the iteration period was 5, and the dropout probability was 50% with the best performance. The graph neural network algorithm based on long and short term preferences had higher recommendation accuracy compared to other algorithms. Modeling users’ short-term and long-term preferences can achieve the goal of comprehensively improving recommendation effectiveness.

REBa2Cu3O7-x (REBCO) coated conductor has emerged as a promising material for the development of ultra-high-field (UHF) magnets. However, the presence of screening-current induced strain within this tape-shaped conductor impacts the operational stability of REBCO coils. We focused on the design of an all-superconducting 36 T / 40 mm UHF magnet, which included a 16 T insert magnet consisting of two nested REBCO coils. In this study, the screening-current induced strain in the 16 T magnet was numerically studied using the coupled electromagnetic-mechanical model, which considered the tilting angle of REBCO tapes and the strain dependency of the critical current. We calculated the screening-current induced strain of each pancake individually and compared these results with those calculated by the sequential model. According to the coupled model, the maximum hoop strain was 0.52%, relatively lower than that calculated by sequential model. The most dangerous pancake was estimated to be the outermost pancake of each coils. Additionally, we varied the critical current value of the REBCO tapes and studied the relationship between the critical current value and screening-current induced strain. This work provides a feasible way to calculate the screening-current effect in large-scale HTS magnets.

Change detection refers to revealing the surface changes from multitemporal images of a given scene, where changed regions of different size and shape may appear anywhere. Convolutional neural network, as one of the most widely used deep learning architectures, has shown good performance in change detection task recently. However, due to multiple convolution and pooling operators in deep architectures, it often happens that small changes are missed and sharp edges are blurred. In this article, an adaptive spatial and difference learning method is proposed to tackle this problem, which mainly contains a feature-adapted difference learning (FADL) module and a difference recalibration (DR) module. In FADL, a kernel-adapted spatial learning part is constructed to capture varying spatial information, which selects proper kernels to adapt different shape and size of changed regions; a joint feature refinement part is employed to learn comprehensive features between bitemporal data, which are further enhanced by difference information. In addition, DR with shallow features is designed to recalibrate spatial and difference details further. From the experiments on three public datasets, our proposed method can relieve the spatial ambiguity problem and obtains the optimal results among the compared change detection methods.

Transformer has shown excellent performance in remote sensing field with long-range modeling capabilities. Remote sensing video (RSV) moving object detection and tracking play indispensable roles in military activities as well as urban monitoring. However, transformers in these fields are still at the exploratory stage. In this survey, we comprehensively summarize the research prospects of transformers in RSV moving object detection and tracking. The core designs of remote sensing transformers and advanced transformers are first analyzed. It mainly includes the attention mechanism evolution for specific tasks, the fitting ability design of input mapping, diverse feature representation, model optimization, etc. The architectural characteristics of RSV detection and tracking are then described across two aspects. One is moving object detection for motion-based traditional background subtractions and appearance-based deep learning models. The other is object tracking for single and multiple targets. The research difficulties mainly include the blurred foreground in RSV data, the irregular object movement in traditional background subtraction, and the severe object occlusion in object tracking. Following that, the potential significance of transformers is discussed according to some thorny problems in RSV. Finally, we summarize ten open challenges of transformers in RSV, which may be used as a reference for promoting future research.

Convolutional neural networks (CNNs) are data-driven methods that automatically extract the rich information embedded in remote sensing images. However, most current deep learning-based remote sensing image change detection methods prioritize high-level semantic features, while not enough attention is given to low-level semantic features, resulting in the loss of edges and details of the change region. To address this problem, this article constructs a spatial-spectral cross fusion network (SSCFNet), divided into the following three modules: 1) a feature extractor network module; 2) a combined enhancement module; 3) a semantic cross-fusion module. A new combined enhancement strategy is proposed to construct several semantic feature blocks in the combined enhancement module. Different convolution operations are applied to the newly constructed semantic feature blocks in the semantic cross fusion module, and the obtained semantic features at various levels are cross-fused. Experiments show that the proposed SSCFNet outperforms the other six state-of-the-art methods on four publicly available remote sensing image change detection datasets.

Brain-inspired algorithms have become a new trend in next-generation artificial intelligence. Through research on brain science, the intelligence of remote sensing algorithms can be effectively improved. This article summarizes and analyzes the essential properties of brain cognize learning and the recent advance of remote sensing interpretation. First, this article introduces the structural composition and the properties of the brain. Then, five represent brain-inspired algorithms are studied, including multiscale geometry analysis, compressed sensing, attention mechanism, reinforcement learning, and transfer learning. Next, this article summarizes the data types of remote sensing, the development of typical applications of remote sensing interpretation, and the implementations of remote sensing, including datasets, software, and hardware. Finally, the top ten open problems and the future direction of brain-inspired remote sensing interpretation are discussed. This work aims to comprehensively review the brain mechanisms and the development of remote sensing and to motivate future research on brain-inspired remote sensing interpretation.

The study of stability criterion and interaction analysis for grid-connected inverter under weak grid is of great value. The previous stability criterion combined current loop, phase-locked loop and other loops together, and used the same external interface to analyze the system stability. It is difficult to independently analyze the interaction among these loops and then design the control loop parameters under the conclusions of the stability analysis. Different from traditional stability criterion, this paper proposes control loop stability criterion for grid-connected inverter from the perspective of controller bandwidth overlap. Firstly, the system d-q overall small-signal model G0 considering phase-locked loop (PLL) of grid-connected inverter under weak grid is given and split into three multiplied independent parts: grid impedance, phase-locked loop and current controller. Then using the equivalent loop ratio expression obtained by combining PLL and grid impedance together and then divided by the current controller, the control loop stability criterion is proposed. The proposed stability criterion can not only maintain the independence of each single loop, but also can analyze the overall system stability. So, further analysis of the interaction law among the three parts under this control loop stability criterion is carried out by deducing the expression of the bandwidth ratio n of the PLL and current controller. And it is found that under the weak grid, the interaction between the links will be generated when the bandwidth ratio is greater than the threshold of n. This phenomenon can be represented by the overlapping area of amplitude-frequency curves in the bode diagram for G0. And the weaker the grid or the closer the bandwidth of PLL is to current controller, the larger the overlapping area of amplitude-frequency curves, and the more likely it is to lead the closed-loop gain to infinity. Furthermore, when the bandwidth of the current controller is fixed, the threshold of n varies in the regions of less than 1 as well as more than 1 with the change of the grid strength. Therefore, the system can remain stable no matter what the bandwidth ratio of the phase-locked loop to the current controller is greater or less than 1. Accuracy of the proposed control loop stability criterion and interaction analysis is verified through simulation and experimental results.

Residual histograms can provide valuable information for vision research. However, current image restoration methods have not fully exploited the potential of multiple residual histograms, especially their role as overall regularization constraints. In this paper, we propose a novel framework of multiple residual Wasserstein driven model (MRWM) that can organically combine multiple residual Wasserstein constraints and various natural image priors for image denoising. Specifically, by utilizing the Wasserstein distance derived from the optimal transmission theory, the multiple residual histograms of the observed images are forced to be as close as possible to the reference residual histogram, thereby improving the accuracy of residual estimation. Furthermore, the proposed concrete MRWM unifies the multiple residual Wasserstein distribution approximation and the image total variation prior knowledge to carry out image denoising. Alternating iterative algorithm of histogram matching and Chambolle dual projection has the characteristics of less parameters and easy implementation. Finally, our experiments confirm that compared with some representative image denoising algorithms, the MRWM can obtain better performance in objective evaluation, and can better preserve the details such as the image edges, making the image look more natural.

Making personalized product integration improvement based on consumer evaluation behavior plays a vital role in enhancing user satisfaction. However, the difference in consumer maturity leads to various standards of satisfaction decision-making in consumer evaluation behavior. Classical theories and models for describing product evaluation decision cannot granularly depict the well-thought-out decision-making process of consumer evaluation behavior and neglect the nonlinear influence of product attributes and consumer personality. Besides, existing studies usually adopt subjective methods like questionnaires and interviews to analyze consumer evaluation behavior, which have a limited number of data and questionable external validity. In order to fill these research gaps, consumer maturity indicators such as product cognition level, preference characteristics and evaluation characteristics are extracted from online reviews, and a three-phase Planned Behavior Model for Consumer Evaluation (PBMCE) is established based on the theory of planned behavior, including the attribute attitude formation phase, the evaluation intention formation phase and the evaluation result generation phase. Simultaneously, a Genetic Algorithm is developed to find the solution of the nonlinear three-phase PBMCE, and personalized product integration improvement strategies are proposed accordingly to achieve high user satisfaction. Finally, experiment results based on 5,200 users’ online reviews validate the effectiveness of the proposed model, indicating that our model can identify the importance of product attributes and consumer personality and characterize their nonlinear impacts on satisfaction decision.

The design, optimization and control methods of a transverse flux Linear Switched Reluctance Motor (LSRM) for the rail transit application are introduced. Two structures of LSRMs, EI type and EE type, are compared and analyzed. The structure of EE type machine is optimized from the perspective of average electromagnetic thrust per unit mass through finite element calculations. The initial sizes of the motor are calculated according to the design theory of the rotary SRM. After obtaining the initial sizes, the sensitivities of the LSRM sizes are analyzed by the three-dimensional finite element calculation, and then the sizes such as the slot depth, the slot width and the tooth width of the motor are optimized. The hardware test platform is built, the electromagnetic characteristics of the prototype are tested, and the motor characteristics of the LSRM with different secondary materials are compared. The control method of the LSRM with position sensors is proposed, and then the sensorless control of the motor is also realized based on the pulse injection strategy. The experimental results verify the effectiveness of the design and control strategies.

Attracting users from a mature large online product community to a new small one by friend recommendation is vital for new product marketing in social network. However, the traditional link prediction algorithms for friend recommendation cannot get high accuracy because of the network sparsity and scale-free problems when attracting customer flow between large and small circles. In order to better adapt to the link prediction of node pairs between circles of different sizes, we propose a collaborative combined link prediction algorithm (CCLPA), which can deeply extract user attention concentration (AC) features in sparse networks. CCLPA possesses three distinctive merits. Firstly, different edges in the network are assigned different attention, and heterogeneous attention concentration indexes (HACIs) within and beyond triadic closure structure are defined accordingly. Second, a random forest (RF) model is designed to adaptively select the appropriate HACIs for a given circle structure, so as to avoid the impact of scale-free problem on link prediction accuracy between different circles. Third, according to the collaboration of the selected indexes and their sensitivity to the circle structure, appropriate sensitive collaborative heterogeneous attention concentration index (SCHACI) is built to avoid the negative impact of blind combination of indexes on predicted performance. Experimental results on Twitter confirm the effectiveness of our proposed method in attracting customer flow in online brand community.

Remote sensing (RS) scene classification plays an essential role in the RS community and has attracted increasing attention due to its wide applications. Recently, benefiting from the powerful feature learning capabilities of convolutional neural networks (CNNs), the accuracy of the RS scene classification has significantly been improved. Although the existing CNN-based methods achieve excellent results, there is still room for improvement. First, the CNN-based methods are adept at capturing the global information from RS scenes. Still, the context relationships hidden in RS scenes cannot be thoroughly mined. Second, due to the specific structure, it is easy for normal CNNs to exploit the heterogenous information from RS scenes. Nevertheless, the homogenous information, which is also crucial to comprehensively understand complex contents within RS scenes, does not get the attention it deserves. Third, most CNNs focus on establishing the relationships between RS scenes and semantic labels. However, the similarities between them are not considered deeply, which are helpful to distinguish the intra-/interclass samples. To overcome the limitations mentioned previously, we propose a homo–heterogenous transformer learning (HHTL) framework for the RS scene classification in this article. First, a patch generation module is designed to generate homogenous and heterogenous patches. Then, a dual-branch feature learning module (FLM) is proposed to mine homogenous and heterogenous information within RS scenes simultaneously. In the FLM, based on vision transformer, not only the global information but also the local areas and their context information can be captured. Finally, we design a classification module, which consists of a fusion submodule and a metric-learning module. It can integrate homo–heterogenous information and compact/separate samples from the same/different RS scene categories. Extensive experiments are conducted on four public RS scene datasets. The encouraging results demonstrate that our HHTL framework can outperform many state-of-the-art methods. Our source codes are available at the below website.

Remote sensing (RS) image scene classification is an important research topic in the RS community, which aims to assign the semantics to the land covers. Recently, due to the strong behavior of convolutional neural network (CNN) in feature representation, the growing number of CNN-based classification methods has been proposed for RS images. Although they achieve cracking performance, there is still some room for improvement. First, apart from the global information, the local features are crucial to distinguish the RS images. The existing networks are good at capturing the global features since the CNNs' hierarchical structure and the nonlinear fitting capacity. However, the local features are not always emphasized. Second, to obtain satisfactory classification results, the distances of RS images from the same/different classes should be minimized/maximized. Nevertheless, these key points in pattern classification do not get the attention they deserve. To overcome the limitation mentioned above, we propose a new CNN named attention consistent network (ACNet) based on the Siamese network in this article. First, due to the dual-branch structure of ACNet, the input data are the image pairs that are obtained by the spatial rotation. This helps our model to fully explore the global features from RS images. Second, we introduce different attention techniques to mine the objects' information from RS images comprehensively. Third, considering the influence of the spatial rotation and the similarities between RS images, we develop an attention consistent model to unify the salient regions and impact/separate the RS images from the same/different semantic categories. Finally, the classification results can be obtained using the learned features. Three popular RS scene datasets are selected to validate our ACNet. Compared with some existing networks, the proposed method can achieve better performance. The encouraging results illustrate that ACNet is effective for the RS image scene classification. The source codes of this method can be found in https://github.com/TangXu-Group/Remote-Sensing-Images-Classification/tree/main/GLCnet.

We present a difference frequency generator (DFG) based on a 2-mm-long MgO-doped aperiodically-poled lithium-niobate (APPLN) crystal with a linearly-varying poling-period. The DFG was pumped by a femtosecond fiber-laser system with a repetition-rate of 53.4 MHz. It produced broadband mid-infrared light source having an instantaneous-bandwidth covering 1.9–4.6 μm, with an average power of 30 mW. Moreover, by numerical simulations, we revealed that due to the relatively high intensity of the incident signal wave, an APPLN crystal with a length of a few milli-meters was sufficient to produce efficient and broadband nonlinear frequency conversion in a high-repetition-rate, ultra-short-pulsed DFG system.

In order to better characterize the performance degradation trend of rolling bearings, a new performance degradation evaluation method based on principal component analysis (PCA), phase space reconstruction (PSR) and kernel extreme learning machine (KELM), namely PAPRKM is proposed to evaluate the performance degradation of rolling bearings in this paper. In the PAPRKM method, the time-domain and frequency-domain features of the vibration signal are extracted to construct the high-dimension feature matrix. Then the PCA is used to reduce the dimension of the feature matrix in order to represent the running state and the declining trend of rolling bearings, so as to eliminate the redundancy and information conflict among these features. Nextly, the PSR is adopted to obtain more relevant information from the time series. By determining the delay time and embedding dimension, the time series are reconstructed to obtain a new performance degradation index, which is regarded as the input data to input into KELM, and the degradation trend prediction model is established to realize the performance degradation trend prediction. Finally, the actual vibration signals of rolling bearings are applied to prove the effectiveness of the PAPRKM. The obtained experimental results show that the PAPRKM method can effectively predict the performance degradation trend of rolling bearings. The predicted results are more accurate than the other compared methods.

With the advent of 5G, multimedia services in wireless communication networks will grow dramatically. So far, many researches have focused on improving security, throughput and channel capacity to support multimedia services. In this paper, a new system model of full-duplex (FD) two-way amplify-and-forward (AF) relay network with simultaneous wireless information and power transfer is presented, where two FD legitimate nodes with two single-antenna communicate with each other via energy-constrained AF relay with multiple antennas, and there is a half-duplex node with single antenna who intends to eavesdrop. This paper targets to maximize the secrecy sum rate of the system by designing the transmit beamforming vector of relay under power constraint and zero-forcing constraint, where two energy harvesting protocols, namely, power splitting and time switching are considered. However the maximization problem is a non-convex optimization, and it is very complex to solve. Therefore a linear method is used to find the lower bound of the secrecy sum rate and the non-convex problem is transformed into a convex optimization problem. After that, an iterative algorithm based on difference-of-convex programming is proposed to solve this convex optimization problem. The simulation results show the good convergence of the proposed iterative algorithm and also the superiority of secrecy sum rate of our proposed scheme compared with two traditional schemes.

Segmentation of pulmonary nodule in thoracic computed tomography (CT) plays an important role in the computer-aided diagnosis (CAD) and clinical practices. However, segmentation of pulmonary nodules still remains a challenging task due to the presence of intrinsic noise, low contrast, intensity-profile inhomogeneity, variable sizes and shapes. Many variants and extensions of fuzzy C-mean (FCM) clustering algorithm have been developed to preserve image details as well as suppress image noises. However, these variants overemphasize the importance of the spatial information and neglect the role of the prior knowledge. To address this problem, a GMM fuzzy C-means (GMMFCM) algorithm is proposed for the segmentation of pulmonary nodules in this paper. A novel local similarity measure is defined by using local spatial information and GMM statistical information. A neighboring term is added to the energy function of traditional fuzzy C-mean algorithm. A superpixel-based random walker is proposed to segment pulmonary parenchyma, which reduces the computational complexity and improves the segmentation performance. Experiments performed on the LIDC dataset and the GHGZMCPLA dataset demonstrate that the segmentation performance of proposed GMMFCM algorithm is superior to the state-of-the-art algorithms.

The existing research demonstrates that the link prediction algorithm which based on community similarity has better prediction performance than that of other node similarity-based methods, and it is more suitable for predicting the probability of the missing links between node-pairs with far distance. However, the disadvantage of these community similarity-based methods is the resolution of prediction accuracy is very low, which resulting in the existence probability of the missing links between node-pairs within a community or between a specific pair of communities is the same. In addition, the link prediction algorithms which based on multi-resolution community division can calculate the existence probability of missing links under different resolutions, but the relevance between communities had not taken into account, which makes it difficult to predict the existence probability of target links if the number of interconnections between communities is small. Combining the advantages of these two algorithms, we propose a more realistic link prediction model which based on a novel quasi-local community relevance index under multi-resolution community division. The performance of our algorithms is demonstrated by comparing with other well-known methods on two kinds of networks in different scales. The experiment results indicate that our approaches are very competitive.

In the Intelligent Transportation Systems (ITS), highly accurate traffic flow prediction is considered as key technology to evaluate traffic state of the urban road network. However, due to disturbing from environment, the original traffic flow data may be influenced by noise and finally cause the decline of prediction accuracy. This study design a hybrid prediction model combining Ensemble Empirical Mode Decomposition (EEMD) denoising schemes and classifying learning algorithm based on Fuzzy C-means Neural Network (FCMNN) to improve prediction accuracy. In the model training process, several key parameters in EEMD and FCMNN are determined according to prediction errors based on traffic volume detected from highway network in the Minneapolis city. In the model validation, three widely used indicators for error evaluation are applied to estimate the prediction accuracy of four candidate models under single and multi step, including Artificial Neural Network (ANN), EEMD+ANN, FCMNN and EEMD+FCMNN. The results shown in the case study indicate that the prediction models combined with denoising methods are superior to the models without adopting denoising algorithm. Furthermore, the model using classifying learning method FCMNN can produce higher prediction accuracy than traditional ANN model. In addition, the long-term prediction performance of FCMNN is also much better than that of ANN because that sub-NN system is trained according to each classifying patterns to obtain better optimization effect. Results summarized in this study could be helpful for administration to design managing and controlling strategies according to high prediction accuracy.

Federated learning (FL) has received considerable attention with the development of mobile internet technology, which is an emerging framework to train a deep learning model from decentralized data. Modern mobile devices often have access to rich but privacy-sensitive data, and computational abilities are often limited because of the hardware restriction. In previous works based on federated averaging (FedAvg) algorithm, mobile devices need to perform lots of calculations, and it is time-consuming in the process of global communication. Inspired by edge computing, we proposed edge federated learning (EdgeFed), which separates the process of updating the local model that is supposed to be completed independently by mobile devices. The outputs of mobile devices are aggregated in the edge server to improve the learning efficiency and decrease the global communication frequency. Empirical experiments demonstrate that our proposed EdgeFed has advantages in different bandwidth scenarios. Especially, by offloading part of the calculations from mobile clients to the edge server, the computational cost of the mobile devices and the global communication expense can be simultaneously reduced as compared to FedAvg.

This paper proposes an optimal generation scheduling approach based on linear relaxation and mixed integer programming, which is used to solve the generation dispatch problem. The quadratic transmission loss constraint of each transmission line is converted into linear constraints by using the linear relaxation and mixed integer programming technique. Consequently, the original optimal generation scheduling problem is formulated as a quadratic programming or mixed integer quadratic programming problem that can be solved by commercial optimization solver. In order to improve the efficiency of algorithm, this paper further analyses the generation scheduling model and deletes the redundant variables and constraints. Three test systems, including IEEE 30-node system, IEEE 118-node system, and Polish 2746-node system, are employed to examine the effectiveness of the proposed method. The comparative results obtained by the proposed method, quadratically constrained quadratic programming method (QCQP), and solving constraint integer programs solver (SCIP) verify the effectiveness of the proposed method in solving the optimal generation scheduling problem.

Polarimetric synthetic aperture radar (PolSAR) image classification has been widely applied in many fields, such as agriculture, meteorology and military. However, some problems, such as the deficiency of labeled data and the underutilization of data information, are always the challenges that can not be ignored in PolSAR image classification. In this paper, a semi-supervised complex-valued generative adversarial networks (SSCV-GANs) is proposed for the first time to address the two issues mentioned above simultaneously. On the one hand, the complex-valued model conforms with the physical mechanism of PolSAR data and it plays an important role for retaining and utilizing amplitude and phase information of PolSAR data. On the other hand, we also present a new complex-valued GANs together with semi-supervised learning to alleviate the problem of insufficient labeled data. Specifically, our complex-valued GANs expands the training data set by generating fake data. Flevoland data and San Francisco data are used to validate the effectiveness of our model. Experimental results show that our model outperforms existing state-of-the-art models in terms of classification accuracy, especially for conditions with fewer labeled data. In particular, the analysis of the statistical distribution of the generated fake data and the real data further demonstrate the effectiveness of the proposed SSCV-GANs.

Ship detection is of considerable significance in both military and civilian application domains. Deep Convolutional Neural Network (DCNN) with region proposal mechanism, e.g., Faster R-CNN, performs outstandingly in ship detection with high-resolution images. However, the accuracy limitation is induced by the region proposal restricted by the training set for multi-scale target detection. Therefore, the method of multi-scale ship object detection is proposed based on saliency estimation in our work. Saliency Estimation Algorithms (SEAs) are often used to extract saliency features in images. In ship detection of high-resolution remote sensing images, these algorithms can extract information such as the scale and position of the targets, and then help the DCNN-based ship detection method to obtain better performance. To verify the effectiveness of the saliency estimation algorithm in multi-scale ship detection of high-resolution remote sensing images, and analyze the advantages of different SEAs. This paper introduces 13 classic saliency estimation algorithms and 2 DCNN-based ones to compare them by using evaluation indicators. At last, in order to demonstrate the performance of different SEAs, the extracted saliency feature maps are used to assist the DCNN-based target detection under multi-scale ships condition. In general, this framework can improve the detection accuracy of large-scale ships under scenarios of training only with small scale ships or without enough datasets.

This article presents a two-stage evolutionary fuzzy clustering framework for noisy image segmentation. It is a bi-stage system comprising a multi-objective optimization stage and a fuzzy clustering segmentation stage. In the multi-objective optimization stage, the fuzzy clustering on pixels in inhomogeneous regions is converted into a multi-objective problem, which can preserve image details while restraining noise. The multi-objective problem is decomposed into several sub-problems by the Tchebycheff approach. A trade-off can be obtained by optimizing these sub-problems simultaneously. In the fuzzy clustering segmentation stage, fuzzy clustering with the trade-off between preserving image details and restraining noise is performed on the whole observed image. To deal with this fuzzy clustering problem, an adaptive evolutionary fuzzy clustering algorithm with spatial information is proposed. Experiment results on synthetic and real images illustrate the effectiveness of the proposed framework for noisy image segmentation.

Synthetic aperture radar (SAR) image classification is an important part in the understanding and interpretation of SAR images. Each patch in SAR images has a scene category, but usually contains multiple land-cover classes or latent properties, which can be represented by topics in the probabilistic topic model (PTM). The representation and selection of discriminative features in PTM have a large impact on the classification results. Most of the existing feature learning methods do not make full use of high-level structure feature and the feature correlation within similar images to mine discriminative features. Therefore, this article proposes a discriminative sketch topic model with structural constraint (C-SSTM) for SAR image classification. In the proposed model, each image patch is characterized by structural and texture features. In particular, the sketch structural feature is based on the sketch map to represent the image local structure pattern. Then, the local image manifold information is preserved in terms of structure and texture. In the structural constraint, the texture and structure of each image patch are combined to learn discriminative latent semantic topics between image patches. Finally, each image patch is quantified by discriminative latent semantic topics instead of low-level representation. The experimental results tested on synthetic and real SAR images demonstrate that the proposed C-SSTM is able to learn effective structural feature representation from SAR images. Compared with other related approaches, C-SSTM produces competitive classification accuracies with high time efficiency.

The general volume of data has exploded to unimaginable levels in the past decade. Therefore, the big data analytics has become an area of focus. Many frameworks have been developed for data analytics, such as Hadoop, Spark, etc. Most of the frameworks are built on multi-core or many-core memory systems, requiring developers and users to have an in-depth understanding of the architectures to take full advantage of the hardware. In this paper, we present a comprehensive study of both multi-core and many-core memory systems and discuss the different characteristics, including core, cache, memory, and the on-chip network. Furthermore, we propose a simple but effective mechanism for cache false-sharing overhead that can reduce a large number of LLC-load/store instructions and LLC cache-misses. In addition, we conduct detailed experiments with Ligra, a graph analytics framework, on four different-sized datasets. The results show that it can achieve up to a 2.5× and 9.5× speed-up for multi-core and many-core memory systems, respectively. We finally share our key findings and discuss the platform development on multi-core and many-core memory systems.

With the resolution increasing, the structure information becomes more and more abundant in Synthetic Aperture Radar (SAR) images. The speckle noise generated by the coherent imaging mechanism, has a great influence on the detection accuracy and detection difficulty accordingly in high-resolution SAR change detection. In this paper, a multivariate change detection framework based on non-subsampled contourlet transform (NSCT), deep belief networks (DBN), fuzzy c-means (FCM) clustering, and global-local spatial pyramid pooling (SPP) net is proposed. NSCT decomposes the image into multiple scales and DBN is used for extracting feature of the decomposed coefficient matrix. FCM converges the similarity matrix of the initial features by DBN into two classes as a pseudo-label for global-local SPP net training data. The global-local SPP net consists of a large-scale region of interest (ROI) SPP net and a small-scale change detection SPP net. The combination of ROI and the SPP net, as well as the fusion between different scales, weakens the interference of the unchanged information and effectively eliminates a large number of redundant information. The experimental results show that our proposed method can effectively remove speckle noise and improve the robustness of high-resolution SAR change detection.

In wireless rechargeable sensor networks for mobile multimedia services, the multi-hop transmission method is often used to reduce the distance of a single hop to save energy. However, the sensor nodes in special locations always carry more forwarding load,which causes energy holes. These sensor nodes in energy hole have the higher energy consumption and the less residual energy, which makes them very fragile and easy to exhaust energy. If a sensor node dies, the structure of the network will be changed and the original function will be affected. In this paper, an energy-balanced joint routing and charging framework is proposed to balance the energy consumption between sensors. By analyzing the network model, the proposed framework is decomposed into a routing strategy and a charging strategy. Considering the impact of the routing structure on sensors energy consumption, the goal of the routing strategy is designed to optimize the routing tree corresponding to the mobile charger staying at each sensor node. Furthermore, the proposed charging strategy prefers the charging path with the best balance of the total energy consumption of sensor nodes during one charging trip. In order to solve these optimization problems, the simulated annealing algorithms, including the proposed random solution generation method and acceptance rules, are used to obtain a fast convergence. The simulation results show that the proposed joint routing and charging framework has better energy balance performance than the traditional scheme.

The Internet of Things (IoT) is an important part of the new generation of information technology. It utilizes many sensor nodes to collect data and monitor the environment and can be applied in various fields. However, because the energy of sensor nodes is limited and the batteries cannot be replaced, the lifetimes of the sensor nodes are limited by the energy of their batteries. Only a reasonable method for saving energy can reduce the energy loss of the nodes in the communication process. Many studies have found that batteries have the property of battery-friendliness; that is, if they continue working, their energy drops rapidly. However, if they rest after working for a period of time, the batteries' energy can be partially restored. In this paper, an effective relay selection scheme for prolonging the lifetimes of nodes is designed based on the battery-friendly nature of the battery. In the proposed battery-friendly relay selection (BFRS) scheme, the relay node is selected based on the weighted synthesis value of the recovered energy, its distance to the sink and the residual energy. A forwarding node that has recovered more energy and is closer to the sink is more likely to be selected as the relay node because of its larger weight. Hence, the BFRS-based routing scheme can improve the lifetime by recovering more energy. Through extensive experiments, we demonstrate that our BFRS scheme outperforms previously proposed schemes. In a network with 500 nodes, compared with the shortest routing algorithm, the network lifetime is increased by 51.26%, and the amount of transferred data is increased by 42.93%.

Extracting knowledge from human mobility data is an important task for many downstream applications such as point-of-interest recommendation, motion trace identification, and personalized trip planning. A specific problem that has recently spurred research interest is the so-called Social Circle Inference from Mobility data (SCIM), aiming at inferring relationships among users based on mobility data and without any explicit structured network information. The existing methods either require partial social ties or fail to model the implicit correlations between user links, thereby suffering from critical inference bias. We present a novel SCIM framework, called SCIM via self-Attention and Contextualized-embedding (SCIMAC) - a methodology capturing multiple aspects of users' check-in behavior and complex motion patterns of different users. Instead of directly applying the recurrent model on training user trajectories, the proposed method introduces a new module for context-aware check-in representation learning by adaptively incorporating the internal states of the recurrent layers, which is more effective than the context-independent check-in embedding used in existing social circle inference approaches. To model the underlying correlations between labels, SCIMAC leverages a more sophisticated label embedding technique to adjust the penalties for correlated users, enabling a better understanding of the user's hierarchy in the label space, and alleviating the inference bias. We empirically demonstrate that our SCIMAC model significantly outperforms several state-of-the-art baselines on real-world datasets.

Air pollution forecasting can provide reliable information about the future pollution situation, which is useful for an efficient operation of air pollution control and helps to plan for prevention. Dynamics of air pollution are usually reflected by various factors, such as the temperature, humidity, wind direction, wind speed, snowfall, rainfall, and so on, which increase the difficulty in understanding the change of air pollutant concentration. In this paper, a short-term forecasting model based on deep learning is proposed for PM2.5 (particulate matter with an aerodynamic diameter less than or equal to $2.5~\mu \text{m}$ ) concentration, and the convolutional-based bidirectional gated recurrent unit (CBGRU) method is presented, which combines 1D convnets (convolutional neural networks) and bidirectional GRU (gated recurrent unit) neural networks. The case is carried out by using the Beijing PM2.5 data set in UCI Machine Learning Repository. Comparing the prediction results with the traditional ones, it is proved that the error of the CBGRU model is lower and the prediction performance is better.

Long Short-Term Memory network (LSTM) is the most widely used recurrent neural network architecture. It plays an important role in a number of research areas, such as language modeling, machine translation and image captioning. However, owing to its recurrent nature, general-purpose processors like CPUs and GPGPUs can achieve limited parallelism while consuming high power energy. FPGA accelerators can outperform general-purpose processors with flexibility, energy-efficiency and more delicate optimization capabilities for the recurrent based algorithms. In this paper, we present the design and implementation of a cloud-oriented FPGA accelerator for LSTM. Different from most of previous works designed for embedded systems, our FPGA accelerator transfers data sequences from and to the host server through PCIe and performs multiple time series predictions in parallel. We optimize both the on-chip computation and the communication between the host server and the FPGA board. We perform experiments to evaluate the overall performance as well as the computation and the PCIe communication efforts. The results show that the performance of our implementation is better than the CPU-based and other hardware-based implementations.

Synthetic aperture radar (SAR) image classification aims at labeling pixels with different categories and this is both, a fundamental step for automatic target recognition (ATR) and a prerequisite for further interpretation. In the past decades, various methods have been proposed for the classification of SAR targets and among them are discriminative dictionary learning (DDL) methods. These DDL methods have recently gained attention from researchers' community due to the fact that they are very powerful on both, representation and discrimination during the classification process of SAR images. However, most of the existing DDL methods adopt l0-norm or l1-norm to ensure the sparsity, but in general, these DDL methods suffer from a high computational burden. Furthermore, it is important to minimize the execution time in the phase of online testing for the scenario of onboard real-time or near real-time SAR automatic target recognition such as modern unmanned aerial vehicle SAR platforms. That said, on reducing execution time, we are confronted with the problem of enhancing recognition efficiency while maintaining its accuracy. In order to solve this problem, our paper proposes a fast DDL method (named as FaDDL) based on a nonlinear analysis co-sparse model by adopting an l1,∞-norm ball as a constraint to replace l0-norm or l1-norm on the coding coefficient matrix. The experimental results show that our proposed method significantly reduces execution time, without losing the classification accuracy.

Deep convolutional neural network (DCNN) can achieve ship detection mission on the high-resolution remote sensing images. However, the false alarms caused by the onshore ship-like objects may decrease the accuracy and feasibility of these DCNN-based detection frameworks. In our work, an end-to-end method, named as Scene Mask R-CNN, is proposed to reduce the onshore false alarms. The scene mask extraction network (SMEN), as a network branch for scene segmentation, is innovatively introduced into the detection framework. The non-target area is marked out by an inferenced scene mask which is used to assist the ship detection. Combining the feature map originated from feature extraction network (FEN) with the inferenced scene mask by using the edge probability weighted (EPW) merging method, the false candidate targets in the non-target area are excluded. This novel mechanism of DCNN-based ship detection not only maintains the detection accuracy, but also effectively suppresses the false alarms in the non-target area. Finally, the validity and accuracy of this method are verified on a ship dataset generated by the high-resolution optical remote sensing images.

Wayside acoustic detection is a promising technology for railway-vehicle bearing health monitoring due to its merits of non-conduct measurement, low cost, and early warning capacity. However, the diagnostic accuracy will be reduced by the problems of strong background noise and Doppler distortion. Considering the super spatial directivity ability of the microphone array, in this paper, a uniform rectangular array (URA) and an optimal spatial filter (OSF) based on the principle of minimum variance distortion-less response (MVDR) are designed to improve the diagnostic accuracy. Compared with the traditional single microphone and linear array, spatial directivity can be improved significantly so that the better anti-noise performance and higher diagnostic accuracy can be achieved. First, a URA consisting of 15 microphone elements arranged into five columns and three rows is designed to capture the wayside acoustic signal. Second, the direction angle of the target moving sound source with high accuracy is calculated at different times. Third, an OSF based on the principle of the MVDR is designed to extract the target sound source signal. Fourth, Doppler effect embedded in the filtered signal is eliminated using the MVDR spectrum estimation and resampling method. Finally, the diagnosis decision is made through an envelope spectrum analysis. The comparative simulation and experimental case studies are carried out to verify the effectiveness and improvement of the proposed method.

With the development of the neural network, some novel reconstructed networks are proposed to solve the problem of compressive sensing (CS) reconstruction. Compare with the traditional reconstruction algorithms, they can reconstruct the original images from the compressed measurement quickly and accurately with a low sampling rate. However, the CS reconstruction algorithms based on neural network ignore the image non-local similarity that is important prior information for the reconstruction. We propose a multi-scale residual reconstruction neural network with non-local constraint (NL_MRN). First, it considers the prior image non-local similarity and adds a non-local operation into the reconstruction network. Then, different scale residual reconstruction modules that have different convolution kernel size are combined to obtain the final output. Finally, the loss function of the whole network is defined as a weighted sum of the loss function of different scale reconstruction modules. What is more, the training efficiency of the network is improved by the proposed segmental training method. The theoretical analysis and the experimental results show that the proposed NL_MRN achieve better reconstruction compared with other reconstruction algorithms, especially at a low sampling rate.

The image segmentation techniques based on multi-level threshold value received lot of attention in recent years. It is because they can be used as a pre-processing step in complex image processing applications. The main problem in identifying the suitable threshold values occurs when classical image segmentation methods are employed. The swarm intelligence (SI) technique is used to improve multi-level threshold image (MTI) segmentation performance. SI technique simulates the social behaviors of swarm ecosystem, such as the behavior exhibited by different birds, animals etc. Based on SI techniques, we developed an alternative MTI segmentation method by using a modified version of the salp swarm algorithm (SSA). The modified algorithm improves the performance of various operators of the moth-flame optimization (MFO) algorithm to address the limitations of traditional SSA algorithm. This results in improved performance of SSA algorithm. In addition, the fuzzy entropy is used as objective function to determine the quality of the solutions. To evaluate the performance of the proposed methodology, we evaluated our techniques on CEC2005 benchmark and Berkeley dataset. Our evaluation results demonstrate that SSAMFO outperforms traditional SSA and MFO algorithms, in terms of PSNR, SSIM and fitness value.

This paper studies a split-complex-valued neuro-fuzzy algorithm for fuzzy inference system, which realizes a frequently used zero-order Takagi-Sugeno-Kang system. Here, adaptive momentum is utilized to speed up the learning convergence. Some strong convergence results are demonstrated based on the weak convergence results, which expresses that the weight sequence of fuzzy parameters converges to a fixed point. Simulation results support the theoretical findings.

Traffic flow prediction is a key step to the efficient operation in the intelligent transportation systems. This paper proposes a hybrid method combing clustering methods and spatiotemporal correlation to predict future traffic trends based on artificial neural network. First, for the traffic flow collected from different loop detectors, a spatio-temporal correlation of data samples is evaluated by considering time correlation and spatial equivalent distance. Second, in order to improve classifying performance and reliability to anomalous data samples, a type-2 fuzzy c-means (FCM) is adopted to make fuzzification of the membership function. Then, a hybrid prediction model combined classification algorithm and neural network is designed to predict various patterns or trends in traffic flow data. Furthermore, the results from the prediction model are modified according to quantized spatio-temporal correlation. Finally, traffic volume data collected from the highway is used to optimize the parameter in the prediction model combination. Several traditional models are used as candidates in comparison, and the higher prediction accuracy demonstrates the effectiveness and feasibility of the hybrid prediction model.