Search

During the past few years, face videos, e.g., video conference, interviews and variety shows, have grown explosively with millions of users over social media networks. Unfortunately, the existing compression algorithms are applied to these videos for reducing bandwidth, which also bring annoying artifacts to face regions. This paper addresses the problem of face quality enhancement in compressed videos by reducing the artifacts of face regions. Specifically, we establish a compressed face video (CFV) database, which includes 196,337 faces in 214 high-quality video sequences and their corresponding 1,712 compressed sequences. We find that the faces of compressed videos exhibit tremendous scale variation and quality fluctuation. Motivated by scalable video coding, we propose a multi-scale recurrent scalable network (MRS-Net+) to enhance the quality of multi-scale faces in compressed videos. The MRS-Net+ is comprised by one base and two refined enhancement levels, corresponding to the quality enhancement of small-, medium- and large-scale faces, respectively. In the multi-level architecture of our MRS-Net+, small-/medium-scale face quality enhancement serves as the basis for facilitating the quality enhancement of medium-/large-scale faces. We further develop a landmark-assisted pyramid alignment (LPA) subnet to align faces across consecutive frames, and then apply the mask-guided quality enhancement (QE) subnet for enhancing multi-scale faces. Finally, experimental results show that our MRS-Net+ method achieves averagely 1.196 dB improvement of peak signal-to-noise ratio (PSNR) and 23.54% saving of Bjontegaard distortion-rate (BD-rate), significantly outperforming other state-of-the-art methods.

In hyperspectral image classification, deep learning (DL) based on abundant training samples has demonstrated its significance in classification performance. However, due to the limitation of available samples and the imbalance/similarity of classes in small-sized datasets, data-driven DL algorithms can hardly extract representative and effective features for interclass classification, and the subtle diagnostic spectral features for intraclass classification are easily covered or lost in the iterative feature extraction (FE). The restricted FE of interclass/intraclass results in the accuracy reduction and performance limitation of refined classification. To mitigate these issues, an attention network with multiscale receptive fields (MRFs) is proposed, embedding an inversion subnet for relative water content retrieval (RWCR). In classification, the three critical parts in the proposed network, namely, MRFs, embedded subnet, and multiple-attention mechanism, are responsible for multiscale feature merging, relative water content (RWC) feature enhancement, and paying attention to bands, channels, and multiscale features, respectively. The ablation studies on small-sized datasets show the accuracy improvements of interclass and intraclass in refined classification, which verifies the effectiveness of critical parts for extracting representative features and taking RWC features as the diagnostic biochemical signature from unbalanced and similar classes. The comparison results with typical DL models demonstrate the superiority of the proposed network. Moreover, the competitive advantage of the proposed network is demonstrated in comparison with traditional and state-of-the-art HSI classification methods.

News recommendation plays an important role in alleviating information overload and helping users find their interesting news. Most of the existing news recommendation methods make a recommendation based on static data. They ignore the time dynamic characteristics of the interaction between users and news, that is, the order in which users click on news implicitly indicates the user’s interest in news. In this paper, we propose a time sensitive heterogeneous graph neural network for news recommendation. The network consists of two subnetworks. One subnet utilizes convolutional neural network and improved LSTM to learn a user’s stay period on the page and click sequence characteristics as the temporal dimension feature. The other subnet constructs an attention-based heterogeneous graph to model the user-news-topic associations, and apply graph neural network to learn the structural features of the heterogeneous graph as spatial dimensional features. Experiments conducted show that our model outperforms the state-of-the-art models in accuracy and has better interpretability.

Utilizing the wavelet theory, the wavelet coefficients with respect to translation and scaling factors can be obtained through the iteration of neural network, which effectively solves the window immobility problem of short time Fourier transform. Notice that, the centralized wavelet neural network is weak in the representation of signal integrity as localized information is lost in the case of the signal properties are large span time-domain and high frequency. To solve this problem, the learning algorithm of distributed wavelet neural networks (DWNNs) based on domain segment is proposed in this paper, where the time frequency characteristic of objective function is set as a constraint condition, while the time width, frequency, center of time domain and band center of the wavelet are employed to determine the translation and scaling parameters. Especially, wavelet networks are distributed into several orthogonal subnet by partitioning the information of input data, and the complexity of calculating for each nodes is thus decreased. Moreover, simulations are presented to demonstrate the advantages and efficiency of the proposed DWNNs.

Currently, several methods applied in distribution systems need power flow solutions, such as evolutionary algorithms, exhaustive search, neural network training, continuation power flow, quasi-static time-series, among others. A properly modeled distribution system requires detailing and expects a three-phase network with neutrals and groundings, which help increase simulation times. For a thorough modeling, but with reduced simulation time, distribution system equivalency can be determined using a multi-area aggregation method in three-phase four-wire multi-grounded feeders, where even the modeling of neutral cables and groundings are allowed. The proposed method enables the user to represent complex network components as a reduced size subnet, achieving a considerable computational time reduction. The proposed method can be inserted into OpenDSS , or multiphase methods, using Newton - Raphson and Backward-Forward Sweep algorithms. Results show that the method is capable of mathematically finding the correct solution for all loading levels. The exactness of the proposition is evaluated using the simple NEV test feeder, IEEE 123, the IEEE 8500 Node Test Feeder, and a meshed network having unbalanced feeders with mutual coupling between different voltage levels.

Layout analysis from a document image plays an important role in document content understanding and information extraction systems. While many existing methods focus on learning knowledge with convolutional networks directly from color channels, we argue the importance of high-frequency structures in document images, especially edge information. In this paper, we present a novel document layout analysis framework with the Explicit Edge Embedding Network (E3 Net ). Specifically, the proposed network contains the edge embedding block and dynamic skip connection block to produce detailed features, as well as a lightweight fully convolutional subnet as the backbone for the effectiveness of the framework. The edge embedding block is designed to explicitly incorporate the edge information from the document images. The dynamic skip connection block aims to learn both color and edge representations with learnable weights. In contrast to the previous methods, we harness the model by using a synthetic document approach to overcome data scarcity. The combination of data augmentation and edge embedding is important toward a more compact representation than directly using the training images with only color channels. We conduct experiments using the proposed framework on three document layout analysis benchmarks and demonstrate its superiority in terms of effectiveness and efficiency over previous approaches.

Visual place recognition (VPR) in environments subject to extreme appearance variation due to changing weather, illumination or seasons is a challenging task. Recent works have shown that features learned from CNNs can achieve promising performance. However, most of the existing methods concentrate so much on the image itself that they neglect the architecture of the network, especially different filters that may carry more meaningful information. In this paper, we develop a learnable feature map filtering (FMF) module constrained by triplet loss to re-calibrate the weight of the individual feature map. In this way, specific feature maps that encode invariant characteristics of location are extracted. Moreover, to make full use of the rich global mutual information that resides in the sample set, we propose an influence-based graph attention network (IB-GAT) with a verification subnet to better incorporate the relations among samples during the training process. Different from conventional GAT approaches, IB-GAT enables feature nodes to attend over the influence of other nodes instead of the original feature. Thus refined features with more discriminative power could be generated. Extensive experiments have been conducted on six public VPR datasets with varying appearances. Ablation analysis verifies the potential efficacy of the FMF module and the IB-GAT components. The experimental results also demonstrate that the proposed methods can achieve better performance than the current state of the art.

Gastric cancer (GC) is the third leading cause of cancer-associated deaths globally. Accurate risk prediction of the overall survival (OS) for GC patients shows significant prognostic value, which helps identify and classify patients into different risk groups to benefit from personalized treatment. Many methods based on machine learning algorithms have been widely explored to predict the risk of OS. However, the accuracy of risk prediction has been limited and remains a challenge with existing methods. Few studies have proposed a framework and pay attention to the low-level and high-level features separately for the risk prediction of OS based on computed tomography images of GC patients. To achieve high accuracy, we propose a multi-focus fusion convolutional neural network. The network focuses on low-level and high-level features, where a subnet to focus on lower-level features and the other enhanced subnet with lateral connection to focus on higher-level semantic features. Three independent datasets of 640 GC patients are used to assess our method. Our proposed network is evaluated by metrics of the concordance index and hazard ratio. Our network outperforms state-of-the-art methods with the highest concordance index and hazard ratio in independent validation and test sets. Our results prove that our architecture can unify the separate low-level and high-level features into a single framework, and can be a powerful method for accurate risk prediction of OS.

The high-frequency single side-band (HF-SSB) speech radio is an essential technique for long-distance speech transmission. However, the HF-SSB received speech is corrupted by both high-power noise and severe channel fading, and the typical speech enhancement methods only focus on the suppression of additive noise. In this letter, a two-stage hybrid approach is proposed to enhance the speech quality of the HF-SSB radio. In the anti-fading stage, we adopt the anti-fading convolution neural network (AF-CNN) to eliminate the effects of channel fading. In the noise suppression stage, noise suppression CNN (NS-CNN) subnet and unsupervised denoising block are used parallelly to further improve the performance and the generalization ability. In addition, we present the optimal topological relations of noise suppression and anti-fading modules by comparison and analysis. Experimental results show that applying the AF-CNN subnet before noise suppression can effectively help recover the weak speech components. Moreover, in terms of objective intelligibility and quality scores, the overall performance of the proposed method outperforms the typical methods that only consider noise suppression.

Most existing urban district heating networks (DHNs) operate on relatively high temperatures. Typically, resulting from 2nd and 3rd generation of district heating (DH), 90–130 °C is used as supply and 50–70 °C as return temperature. On the contrary, low temperature district heating networks (LTDHN) with supply temperatures between 30 °C and 70 °C allow a more efficient operation of the network and can further utilize renewable heat sources such as solar or geothermal energy, low temperature waste heat, or even ambient heat. However, high temperature DHNs cannot easily be modified and transformed into LTDHNs due to established operating conditions and assets. Therefore, the integration of a low-temperature sub-network (sub-LTDHN) and creating an energy cascade is an opportunity to increase the efficiency and sustainability of the overall network. The sub-LTDHN is integrated into the return line of the superior DHN, hence leading to a reduction of the return temperature of the superior DHN. Sub-LTDHNs can be a key enabler for the decarbonization of urban DHNs by enabling an efficient utilization of local energy sources and have the potential to reduce substantially the overall network temperatures. At the same time, a high temperature backup from the main network to the sub-LTDHN is available. Within this study, theoretical conceptualized as well as practical examples of sub-LTDHN are compiled using literature research, expert interviews and the professional international network to DHN operators. Despite the potential advantages of sub-LTDHNs, there are several technical, legislative and economic issues related to their implementation, which is the reason why they are hardly implemented. The review investigates the revealed solutions and conditions of the compiled sub-LTDHNs and tries to generalize them in order to allow its replicability. Therefore, general results on implementations, frameworks, barriers and enablers are derived (“lessons learnt”) and good practices are properly elaborated.

The awareness of physical network topology in a large-scale Internet of Things (IoT) system is critical to enable location-based service provisioning and performance optimization. However, due to the dynamics and complexity of IoT networks, it is usually very difficult to discover and update the physical topology of the large-scale IoT systems in real-time. Considering the stringent latency requirements in IoT systems, while the initial processing time for topology discovery can be tolerated, latency due to real-time topology update constitutes an even higher level of challenge. In this paper, a novel fast hierarchical topology update scheme is proposed for the large-scale IoT systems enabled by using the edge-cloud collaborative architecture. Specifically, an event-driven neighbor update algorithm, termed as TriggerOn, is firstly developed to update the local neighbor table of the end devices when device association or disassociation occurs. Based on the updated neighbor tables, the physical topology update of the subnet is conducted at the coordinated edge device, where a hybrid multidimensional scaling (MDS) based 3D localization algorithm is developed to locate the newly associated devices. Simulation results have indicated that as compared to the benchmark methods, the neighbor discovery latency has been reduced dramatically, and the 3D localization accuracy has been improved. Furthermore, the overall latency incurred by the proposed hierarchical physical topology update scheme is significantly lower than the distributed consensus-based update scheme, especially for the large-scale IoT subnets.

Software defined satellite-terrestrial networking has been identified as a promising approach to support the diversity of network services. As the fundamental issue to improve the flexibility of network management, the controller placement problem has been attracted increasing attentions for the integration of satellite and terrestrial networking. However, the impact of the dynamic coverage demands on the controller placement have not been well investigated in existed works, which makes them fail to adjust the coverage dynamically according to the actual demands, and leads to an obvious increase of networking response latency to the terminals. Aiming to address this issue, we propose a novel on-demand dynamic controller placement scheme, which can optimize the placement of controllers to improve networking response latency while meeting the dynamic coverage demands. Firstly, to optimize the number of controllers and meet the dynamic coverage demands, we define the coverage redundancy and propose the redundancy-based satellite subnet division method to establish the reliable satellite subnets. Secondly, we quantify the networking response latency of the distributed satellite subnet, and build an optimization mathematical model to optimize the number and location of controllers. Then, we formulate the controller placement problem into the capacitated facility location problem and build the mathematical model for it. Moreover, the on-demand dynamic approximation algorithm is proposed to obtain the approximation solution. Finally, the simulation results demonstrate that the proposed algorithm can effectively optimize the network latency compared with related algorithms.

Given the outbreak of COVID-19 pandemic and the shortage of medical resource, extensive deep learning models have been proposed for automatic COVID-19 diagnosis, based on 3D computed tomography (CT) scans. However, the existing models independently process the 3D lesion segmentation and disease classification, ignoring the inherent correlation between these two tasks. In this paper, we propose a joint deep learning model of 3D lesion segmentation and classification for diagnosing COVID-19, called DeepSC-COVID, as the first attempt in this direction. Specifically, we establish a large-scale CT database containing 1,805 3D CT scans with fine-grained lesion annotations, and reveal 4 findings about lesion difference between COVID-19 and community acquired pneumonia (CAP). Inspired by our findings, DeepSC-COVID is designed with 3 subnets: a cross-task feature subnet for feature extraction, a 3D lesion subnet for lesion segmentation, and a classification subnet for disease diagnosis. Besides, the task-aware loss is proposed for learning the task interaction across the 3D lesion and classification subnets. Different from all existing models for COVID-19 diagnosis, our model is interpretable with fine-grained 3D lesion distribution. Finally, extensive experimental results show that the joint learning framework in our model significantly improves the performance of 3D lesion segmentation and disease classification in both efficiency and efficacy.

Face age progression, goals to alter the individual's face from a given face image to predict the future appearance of that image. In today's world that demands more security and a touchless unique identification system, face aging attains tremendous attention. The existing face age progression approaches have the key problem of unnatural modifications of facial attributes due to insufficient prior knowledge of input images and nearly visual artifacts in the generated output. Research has been continuing in face aging to handle the challenge to generate aged faces accurately. So, to solve the issue, the proposed work focuses on the realistic face aging method using AttentionGAN and SRGAN. AttentionGAN uses two separate subnets in a generator. One subnet for generating multiple attention masks and the other for generating multiple content masks. Then attention mask is multiplied with the corresponding content mask along with an input image to finally achieve the desired results. Further, the regex filtering process is performed to separates the synthesized face images from the output of AttentionGAN. Then image sharpening with edge enhancement is done to give high-quality input to SRGAN, which further generates the super-resolution face aged images. Thus, presents more detailed information in an image because of its high quality. Moreover, the experimental results are obtained from five publicly available datasets: UTKFace, CACD, FGNET, IMDB-WIKI, and CelebA. The proposed work is evaluated with quantitative and qualitative methods, produces synthesized face aged images with a 0.001% error rate, and is also evaluated with the comparison to prior methods. The paper focuses on the various practical applications of super-resolution face aging using Generative Adversarial Networks (GANs).

In recent years, cloud computing is becoming an attractive research topic for its emerging issues and challenges. Not only in research but also the enterprises are rapidly adopting cloud computing because of its numerous profitable services. Cloud computing provides a variety of quality of services (QoSs) and allows its users to access these services in the form of infrastructure, platform and software on a subscription basis. However, due to its flexible nature and huge benefits, the demand for cloud computing is rising day by day. As a circumstance, many cloud service providers (CSPs) have been providing services in the cloud market. Therefore, it becomes significantly cumbersome for cloud users to select an appropriate CSP, especially considering various QoS criteria. This paper presents a hybrid multi-criteria decision-making (H-MCDM) algorithm to find a solution by considering different conflicting QoS criteria. The proposed algorithm takes advantage of two well-known MCDM algorithms, namely analytic network process (ANP) and VIseKriterijumska Optimizacija I Kompromisno Resenje (VIKOR), to select the best CSP or alternative. Here, ANP is used to categorize the criteria into subnets and finds the local rank of the CSPs in each subnet, followed by VIKOR, to find the global rank of the CSPs. H-MCDM considers both beneficial and non-beneficial criteria and finds the CSP that holds the maximum and minimum values of these criteria, respectively. We demonstrate the performance of H-MCDM using a real-life test case (case study) and compare the results to show the efficacy. Finally, we perform a sensitivity analysis to show the robustness and stability of our algorithm.

Multi-dimensional data exploration is a classic research topic in visualization. Most existing approaches are designed for identifying record patterns in dimensional space or subspace. In this paper, we propose a visual analytics approach to exploring subset patterns. The core of the approach is a subset embedding network (SEN) that represents a group of subsets as uniformly-formatted embeddings. We implement the SEN as multiple subnets with separate loss functions. The design enables to handle arbitrary subsets and capture the similarity of subsets on single features, thus achieving accurate pattern exploration, which in most cases is searching for subsets having similar values on few features. Moreover, each subnet is a fully-connected neural network with one hidden layer. The simple structure brings high training efficiency. We integrate the SEN into a visualization system that achieves a 3-step workflow. Specifically, analysts (1) partition the given dataset into subsets, (2) select portions in a projected latent space created using the SEN, and (3) determine the existence of patterns within selected subsets. Generally, the system combines visualizations, interactions, automatic methods, and quantitative measures to balance the exploration flexibility and operation efficiency, and improve the interpretability and faithfulness of the identified patterns. Case studies and quantitative experiments on multiple open datasets demonstrate the general applicability and effectiveness of our approach.

Drawing on complex network theory, this paper combines bus network and subway network into a multi-subnet composited complex network (MSCCN) of urban public transport (UPT). Then, the cascading failure of the MSCCN nodes and edges was modelled, and the passenger flow transfer rules were established under node and edge failures. Considering the impact of extreme weather on the UPT, a synthetic operator was created to measure the robustness of UPT MSCCN, in the light of network topology and passenger flow, and used to quantify the robustness of UPT MSCCN under extreme weather. Finally, a UPT MSCCN was set up for Qingdao, China, and subject to cascading failure simulation. The simulation results reveal how MSCCN robustness changes with different variables. The research findings help to optimize the response of the MSCCN to extreme weather.

Embedded systems have numerous applications in everyday life. Petri-net-based representation for embedded systems (PRES+) is an important methodology for the modeling and analysis of these embedded systems. For a large complex embedded system, the state space explosion is a difficult problem for PRES+ to model and analyze. The Petri net synthesis method allows one to bypass the state space explosion issue. To solve this problem, as well as model and analyze large complex systems, two synthesis methods for PRES+ are presented in this paper. First, the property preservation of the synthesis shared transition set method is investigated. The property preservation of the synthesis shared transition subnet set method is then studied. An abstraction-synthesis-refinement representation method is proposed. Through this representation method, the synthesis shared transition set approach is used to investigate the property preservation of the synthesis shared transition subnet set operation. Under certain conditions, several important properties of these synthetic nets are preserved, namely reachability, timing, functionality, and liveness. An embedded control system model is used as an example to illustrate the effectiveness of these synthesis methods for PRES+.

Domain Name System (DNS) is a critical service for enterprise operations, and is often made openly accessible across firewalls. Malicious actors use this fact to attack organizational DNS servers, or use them as reflectors to attack other victims. Further, attackers can operate with little resources, can hide behind open recursive resolvers, and can amplify their attack volume manifold. The rising frequency and effectiveness of DNS-based DDoS attacks make this a growing concern for organizations. Solutions available today, such as firewalls and intrusion detection systems, use combinations of black-lists of malicious sources and thresholds on DNS traffic volumes to detect and defend against volumetric attacks, which are not robust to attack sources that morph their identity or adapt their rates to evade detection. We propose a method for detecting distributed DNS attacks that uses a hierarchical graph structure to track DNS traffic at three levels of host, subnet, and autonomous system (AS), combined with machine learning that identifies anomalous behaviors at various levels of the hierarchy. Our method can detect distributed attacks even with low rates and stealthy patterns. Our contributions are three-fold: (1) We analyze real DNS traffic over a week (nearly 400M packets) from the edges of two large enterprise networks to highlight various types of incoming DNS queries and the behavior of malicious entities generating query scans and floods; (2) We develop a hierarchical graph structure to monitor DNS activity, identify key attributes, and train/tune/evaluate anomaly detection models for various levels of the hierarchy, yielding more than 99% accuracy at each level; and (3) We apply our scheme to a month’s worth of DNS data from the two enterprises and compare the results against blacklists and firewall logs to demonstrate its ability in detecting distributed attacks that might be missed by legacy methods while maintaining a decent real-time performance.

The spiking neural network (SNN) is considered to be the third generation of neural networks featured by its low power consumption and high computing capability, which has great application potential in robotics. However, the present SNN has two limitations: 1) the neuron’s spike firing time is calculated based on the iterative approach, which dramatically slows down the calculation rate of the SNN and 2) the existing learning algorithm is more suitable for the single-layer structure, which can hardly train the network with “deep structure.” To this end, this paper proposes a novel spike firing time search algorithm that can narrow the search interval. In addition, a pretrained subnet SNN is designed, which makes the SNN have more hidden layers. This setting of the SNN can effectively improve its performance in pattern recognition tasks. Furthermore, by using the surface electromyography signal (sEMG), the proposed SNN is used to recognize the hand gestures. The experimental results show that: 1) the spike firing time search algorithm can significantly increase the forward propagation rate of the SNN and 2) the proposed SNN can reach a satisfactory recognition accuracy ratio 97.4%, which is 0.9% higher than that of the fully connected SNN.

Proactive defense is one of the most promising approaches to enhance cyber-security in the power systems, while how to balance its costs and benefits has not been fully studied. This paper proposes a novel method to model cyber adversarial behaviors as attackers contending for the defenders’ benefit based on the game theory. We firstly calculate the final benefit of the hackers and defenders in different states on the basis of the constructed models and then predict the possible attack behavior and evaluate the best defense strategy for the power systems. Based on a real power system subnet, we analyze 27 attack models with our method, and the result shows that the optimal strategy of the attacker is to launch a small-scale attack. Correspondingly, the optimal strategy of the defender is to conduct partial-defense.

This paper presents a lightweight Multi-task Semantic Attention Network (MTSAN) to collectively deal with object detection as well as semantic segmentation aiding real-time applications of the Advanced Driver Assistance Systems (ADAS). This paper proposes a Semantic Attention Module (SAM) that introduces the semantic contextual clues from a segmentation subnet to guide a detection subnet. The SAM significantly boosts up the detection performance and computational cost by considerably decreasing the false alarm rate and it is completely independent of any other parameters. The experimental results show the effectiveness of each component of the network and demonstrate that the proposed MTSAN yields a better balance between accuracy and speed. Following the post-processing methods, the proposed module is tested and proved for its accuracy in the Lane Departure Warning System (LDWS) and Forward Collision Warning System (FCWS). In addition, the proposed lightweight network is deployable on low-power embedded devices to meet the requirements of the real-time applications yielding 10FPS @ 512 X 256 on NVIDIA Jetson Xavier and 15FPS @ 512 X 256 on Texas Instrument’s TDA2x.

Detecting defects, which is a branch of target detection in the field of computer vision, is widely used in factory production. To solve the problems in existing detection algorithms that relate to their insensitivity to large or medium defect targets on bearing covers, their difficulty in detecting subtle defects effectively and their lack of real-time detection, in this work, we establish a large-scale bearing-cover defect dataset and propose an improved YOLOv3 network model. The proposed model is divided into four submodels: the bottleneck attention network (BNA-Net), the attention prediction subnet model, the defect localization subnet model, and the large-size output feature branch. To test the generality, robustness and practicability of the new model, we design a comparative experiment under abnormal illumination conditions. We design an ablation experiment to verify the validity of the proposed submodules. The experimental results show that our model solves the problem of the YOLOv3 algorithm’s insensitivity to medium or large targets and satisfies real-time detection conditions. The mAP result is 69.74%, which is 16.31%, 13.4%, 13%, 10.9%, and 7.2% more than that of YOLOv3, EfficientDet-D2, YOLOv5, YOLOv4, and PP-YOLO, respectively.

A reconfigurable manufacturing system (RMS) means that it can be reconfigured and become more complex during its operation. In RMSs, deadlocks may occur because of sharing of reliable or unreliable resources. Various deadlock control techniques are proposed for RMSs with reliable and unreliable resources. However, when the system is large-sized, the complexity of these techniques will increase. To overcome this problem, this paper develops a four-step deadlock control policy for the detection and treatment of faults in an RMS. In the first step, a colored resource-oriented timed Petri net (CROTPN) is designed for rapid and effective reconfiguration of the RMS without considering resource failures. In the second step, “sufficient and necessary conditions” for the liveness of a CROTPN are introduced to guarantee that the model is live. The third step considers the problems of failures of all resources in the CROTPN model and guarantees that the model is reliable by designing a common recovery subnet and adding it to the obtained CROTPN model at the second step. The fourth step designs a new hybrid method that combines the CROTPN with neural networks for fault detection and treatment. A simulation is performed using the GPenSIM tool to evaluate the proposed policy under the RMS configuration changes and the results are compared with the existing approaches in the literature. It is shown that the proposed approach can handle any complex RMS configurations, solve the deadlock problem in an RMS, and detect and treat failures. Furthermore, is simpler in its structure.

Single image deraining is a fundamental pre-processing step in many computer vision applications for improving the visual effect and system performance of the downstream high-level tasks in adverse weather conditions. This study proposes a novel multi-scale context aggregation network, to effectively solve the single image deraining problem. Specifically, we exploit a lightweight residual structure subnet as the baseline architecture to extract fine and detailed texture context at the original scale and further incorporate a multi-scale progressive aggregation module (MPAM) to learn the complementary high-level context for enhancing the modeling capability of the overall deraining network. The MPAM, designed as a plug-and-play module to be utilized in the arbitrary network, is composed of multi-scale convolution blocks to learn a wide variety of contexts in multiple receptive fields, and then carries out progressive context aggregation between adjacent scales with residual connections, which is expected to concurrently disentangle the multi-scale structures of scene contents and multiple rain layers in the rainy images, and models more representative contexts for reconstructing the clean image. To reduce the learnable parameters in the MPAM, we further explore a context hallucinate block for replacing the multi-scale convolution block, and propose a lightweight MPAM. Moreover, for being specially adaptive to deal with the input rainy images with a lot of unwanted components (rain layers), we delve into the artifact-attenuating pooling and activation functions via taking into consideration of the surrounding spatial context instead of pixel-wise operation and propose the spatial context-aware pooling (SCAP) and activation (SCAA) for incorporating with our deraining network to boost performance. Extensive experiments on the benchmark datasets demonstrate that our proposed method performs favorably against state-of-the-art deraining approaches.

The context-based convolutional neural network (CNN) is one of the most well-known CNNs to improve the performance of semantic segmentation. It has achieved remarkable success in various medical image segmentation tasks. However, extracting rich and useful context information from complex and changeable medical images is a challenge for medical image segmentation. In this study, a novel Multi-Receptive-Field CNN (MRFNet) is proposed to tackle this challenge. MRFNet offers the optimal receptive field for each subnet in the encoder-decoder module (EDM) and generates multi-receptive-field context information at the feature map level. Moreover, MRFNet fuses these multi-feature maps by the concatenation operation. MRFNet is evaluated on 3 public medical image data sets, including SISS, 3DIRCADb, and SPES. Experimental results show that MRFNet achieves the outstanding performance on all 3 data sets, and outperforms other segmentation methods on 3DIRCADb test set without pre-training the model.

A net of intermolecular contacts in organic crystals contains a finite number of strong contacts sufficient for the net formation. These contacts constitute a subnet referred to as the critical net. Critical nets of 1754 crystalline hydrocarbons in six most common structural classes (supergiants) are studied. Coordination numbers and the critical net topology are analyzed, the predominant topological types (dia, pcu, hex, bnn, etc.) are revealed. The critical net is found to be quite economical in 95% cases and containing n + Δ (Δ = 0-2) symmetrically independent contacts, where n is their minimal possible number.

Metro passenger detection is always a significant task and a bottleneck in metro video surveillance system. Much recent research has demonstrated that Convolutional Neural Network (CNN) is more powerful than other machine learning algorithms in numerous computer vision tasks. Motivated by the research, this paper proposes MetroNet and Tiny MetroNet for detecting occluded metro passengers in metro embedded system with limited hardware resources. MetroNet consists of smaller CNN-SqueezeNet, Region Proposal Network (RPN) and Detection Head subnet. Besides, the repulsion loss is adopted to effectively prevent detection results from worsening caused by severe passengers’ occlusion during training phase. On the other hand, considering that some platforms have more limited hardware resources, a simple version of the MetroNet named Tiny MetroNet is designed and a novel, tiny passenger feature network is proposed as backbone. Based on three datasets, two MetroNets are tested and compared to existing state-of-the-art detection networks on CPU and GPU mode. The experiment results demonstrate that MetroNet has real-time performance and better detection accuracy. Tiny MetroNet achieves fast detection speed and smaller model size with acceptable performance degradation. Even for the ARM embedded system, their performance is competitive and can meet the application requirements of high-speed metros.

Deep neural networks have recently demonstrated high performance for deblurring. However, few methods are designed for both non-uniform image blur estimation and removal with highly efficient. In this study, the authors proposed a fully convolutional network that outputs estimated blur and restored image in one feed-forward pass for the non-uniformly blurred image of any input-size. The proposed network contains two subnets. The parameter estimation subnet P-net predicts pixel-wise parameters of multiple blur types with high accuracy. The output of P-net is used as a condition, which guides the blur removal subnet G-net to restore a high quality latent sharp image. P-net and G-net are ultimately integrated into a single framework called PG-net, which guarantees the consistency of parameter estimation and blur removal, thereby improves algorithm efficiency. Experiment results show that the authors blur parameter estimation method as well as their deblurring method outperforms the comparison methods both quantitatively and qualitatively.

Calculation of subnetting IP Address manually is quite time consuming and difficult for people who are just learning. With current smartphone technology, especially those using the Android operating system. There are many things that can be done with smartphones nowadays, including studying subnetting IP address calculations. The purpose of this study is to design and build a mobile application to facilitate the study of subnetting IP Address calculations with the Android operating system. The research method used is a waterfall with stages of requirements, design, implementation, and testing. This research resulted in a mobile application for calculating IP Address subnetting that will facilitate studying and accelerating the calculation of IP Address subnetting. This application is able to search for subnet masks, number of subnets, number of hosts per subnet, IP broadcast, IP range, network ID, and there is a theoretical explanation of the IP Address, prefix, and range of each class of IP Address. The results of the study are indicated by the level of eligibility of this application based on a questionnaire from users with results, 41% strongly agree, 44% agree, 13% neutral, and 2% disagree.Keywords: IP Address, Subneting, Application, Android.

Voltage-dependent resistors (VDRs) are important circuit-protection devices. Their performance is affected by packaging quality. To identify VDR packaging defects more accurately and efficiently, we have proposed a convolutional neural network (CNN)-based VDR appearance quality inspection method that includes four stages: image acquisition, data augmentation, neural architecture design, and CNN training and testing. In designing the neural architecture, we have proposed two VDR-oriented network blocks, which consist of a compressed subnet and a multiscale subnet. Then, a stacking-block-based neural architecture design (BlockNAD) strategy is employed to determine the number of blocks. The last block is connected to a classification layer composed of a global average pooling (GAP) layer and a full connection (FC) layer. Further, using a VDR dataset containing 8058 images, we compared the identification performances of the candidate networks with different structures on 12 categories of VDR defects by adopting a variety of indicators, such as the mean average precision (mAP) and average test time per sample. The experimental results of the proposed method demonstrate competitive results compared to the state-of-the-art methods in identifying VDR defects, with a mAP value of approximately 99.9% and an average test time per sample of approximately 3 ms.

The paper investigates the use of generative adversarial networks (GAN) for intentional modification of Earth remote sensing data. A generative neural network that includes a special subnet for object boundary inpainting is considered. The network comprises two GAN: the first completes the object boundary, and the second repaints blank areas. Actual remote sensing data are used to test the generative network under consideration. The exemplar-based Patch-Match algorithm is taken as a reference for comparison purposes. The experimental results allow the conclusion that the approach is an effective tool for the intentional modification of large terrestrial area images in falsification of Earth remote sensing data.

An adaptive routing helps in evading early network saturation by steering data packets through the less congested area at the oppressive loaded situation. However, performances of adaptive routing are not always promising under all circumstances. Say for, given more freedom in choosing an alternate route on non-minimal paths for a substantially loaded network even may result in worsening network performances due to following longer route under adaptive routing. Here, underlying topology facilitates routing by offering more alternate short-cut routes on minimal or quasi-minimal paths. This work presents a congestion-aware (CA) adaptive routing for one-hop diagonally connected subnet-based mesh (SDmesh) network aiming to facilitate both performances and routing flexibility simultaneously. Our proposed technique on the selected system facilitates packet routing, offering more options in choosing an output link from minimal or quasi-minimal paths and hence helps in lowering packet delay by shortening the length of traversed traffic under the oppressive loaded situation. Furthermore, we have also employed a congestion-aware virtual input crossbar router aiming to split the entire network into two distinct logically separated sub-networks. It facilitates preserving important routing properties like deadlock, live-lock fairness, and other essential routing constraints. Experiments, conducted over two 8×8- and 12×12-sized networks, show an average improvement of 25--87.5% saturated latency and 60--83% throughput improvement under uniform traffic patterns for the proposed CA routing compared to centralized adaptive XY routing. Experimental results on application-specific PARSEC and SPLASH2 benchmark suites show an average of 22--50% latency and 23--30% throughput improvements by the proposed technique compared to centralized XY routing on the baseline mesh network. Moreover, experiments were also carried out to check the performance of the proposed routing method with different newly proposed deadlock-free adaptive routing approaches over the same subnet-based diagonal mesh (SDmesh) network and reported.

In this paper, based on the exponential synchronization of double-layered networks composed of multiple subnets, we bring forward a chaotic secure communication scheme, which is robust to different network sizes, time-varying delays, and stochastic noise. Cluster analysis is used to process the nodes with different dynamics, such that the subnets in the transmitter and receiver are one-to-one correspondence and constitute subnet pairs, while the node size of each subnet can be inconsistent. Each subnet pair is only accountable for encrypting/decrypting a part of information. There are many encryption/decryption units and they are operating in parallel, which can not only indicate the complex characteristic of different nodes but also speed up the information encryption/decryption. The proposed scheme applies the chaotic signals yielded by many chaotic systems as key sequences, such that the key space can grow with node sizes in the transmitter. This scheme is not subject to the constraint on the amplitude of the original message and chaotic signal, the influence of time-delay and stochastic noise, and it shows that only the nodes with large degrees need to apply controllers in restoring the original information. Both theoretical analysis and numerical simulation manifest the feasibility and validity of the given scheme.

Over the past two decades, the research community has developed many approaches to study the Internet topology. In particular, starting from 2007, various tools explored the inference of subnets, i.e., sets of devices located on the same connection medium which can communicate directly with each other at the link layer. In this paper, we first discuss how today’s traffic engineering policies increase the difficulty of subnet inference. We carefully characterize typical difficulties and quantify them in the wild. Next, we introduce WISE (Wide and lInear Subnet inferencE), a new tool which tackles those difficulties and discovers, in a linear time, large networks subnets. Based on two ground truth networks, we demonstrate that WISE outperforms state-of-the-art tools. Then, through large-scale measurements, we show that the selection of a vantage point with WISE has a marginal effect regarding accuracy. Finally, we discuss how subnets can be used to infer neighborhoods (i.e., aggregates of subnets located at most one hop from each other). We discuss how these neighborhoods can lead to bipartite models of the Internet and present validation results and an evaluation of neighborhoods in the wild, using WISE . Both our code and data are freely available.

In order to improve CS performance of natural images, in this paper, we propose a novel framework to design an OPtimization-INspired Explicable deep Network, dubbed OPINE-Net, for adaptive sampling and recovery. Both orthogonal and binary constraints of sampling matrix are incorporated into OPINE-Net simultaneously. In particular, OPINE-Net is composed of three subnets: sampling subnet, initialization subnet and recovery subnet, and all the parameters in OPINE-Net (e.g. sampling matrix, nonlinear transforms, shrinkage threshold) are learned end-to-end, rather than hand-crafted. Moreover, considering the relationship among neighboring blocks, an enhanced version OPINE-Net $^+$ is developed, which allows image blocks to be sampled independently but reconstructed jointly to further enhance the performance. In addition, some interesting findings of learned sampling matrix are presented. Compared with existing state-of-the-art network-based CS methods, the proposed hardware-friendly OPINE-Nets not only achieve better performance but also require much fewer parameters and much less storage space, while maintaining a real-time running speed.

In this paper, a general description of the problem of calculating probability–time characteristics was presented. At the initial stage of studying this problem, we studied the asynchronous data transmission of the constituent broadband digital network with the integration of services for both the channel switching subnetwork and the packet switching subnetwork. And also the circuit switching subnetwork was investigated using the bypass of the direction of transmission of the stream of multi-channel calls. At the same time, the source data of the problem of calculating probabilistic characteristics were set for the channel switching subnetwork, the characteristics that were determined during the solution of the problem were listed, and some assumptions were described that made it possible to adequately approximate the study of such a model to the functioning of a real network. Next, a mathematical model for calculating the loss probabilities was described, both for communication channels and for nodes of the channel switching subnetwork included in the asynchronous transfer mode network. The model of information traffic transmission is presented, where the method is implemented using network bypass directions. A method has been developed for generating a nodal load of a channel switching subnet and a system of equations has been obtained, the solution of which determines the value of a nodal load relative to the probability of loss on communication channels and proved the uniqueness of the solution of this system of equations.

Distributed network architecture of heterogeneous computing faces with such problems as strict performance constraints of network control, unpredictable mapping relationship between computing data algorithms of different mobile terminals and inconsistency between computing algorithms and link control of data networks. In order to solve the above problems, we begin with software definition network architecture and load balancing algorithm for heterogeneous computing, and gradually improve the real-time and reliability of heterogeneous computing. On the one hand, the heterogeneous computing data of fog node and cloud computing system are distributed. The centralized service of software-defined network combines with distributed computing of mobile edge terminal and its subnet. On the other hand, we define the centralized information and distributed scheduler of the network. In addition, we deploy the optimal assignment of data sharing and heterogeneous computing tasks in real time with ellipse-partitioned area as the object. A series of algorithms for classifying and assigning heterogeneous computing data streams in software-defined networks are designed to achieve the optimal balance among load balancing, minimum classification of large data streams, minimum resource occupation and time constraints. Experimental comparison compared and evaluated the Load Balancing with big data stream (LBBS), Load Balancing with Heterogeneous Computing (LBHC) and the proposed LBBHD. Compared with the other two algorithms, the proposed algorithm improves workload skewness, throughput and load balancing error respectively about 2.1%, 1.96%, 2.9%, 2.2%; 5.57%. 2.51%.