Search

Network measurement provides operators an efficient tool for many network management tasks such as performance diagnosis, traffic engineering and intrusion prevention. However, with the rapid and continuous growth of traffic speed, it needs more computing and memory resources to monitor traffic in per-flow or per-packet granularity. Sample-based measurement systems (e.g., NetFlow, sFlow) have been developed to perform coarse-grained measurement, but they may miss part of records, especially for mice flows, which are important for some network management tasks (e.g., anomaly detection, performance diagnosis). To address these issues, data streaming algorithms such as hash tables and sketches have been introduced to balance the trade-off among accuracy, speed, and memory usage. In this paper, we present a systematic survey of various data structures, algorithms and systems which have been proposed in recent years to perform fine-grained measurement for high-speed networks. We organize these methods and systems from a software-defined perspective. In particular, we abstract fine-grained network measurement into three-layer architecture. We introduce the responsibility of each layer and categorize existing state-of-the-art works into this architecture. Finally, we conclude the paper and discuss the future directions of fine-grained network measurement.

With the rapid development of Internet of Things (IoT), digital twin is emerging as one of the most promising technologies to connect physical components with digital space for better optimization of physical systems. However, the limited wireless resource and security concerns impede the deployment of digital twin in IoT. In this paper, we exploit blockchain to propose a new digital twin edge networks framework for enabling flexible and secure digital twin construction. We first develop cooperative federated learning through access point (AP) to help resource limited smart devices in constructing digital twin at the network edges belonging to different mobile network operators (MNOs). Then, we propose a model update chain by leveraging Directed Acyclic Graph (DAG) blockchain to secure both local model updates and global model updates. In order to incentivize the APs to help in local models training for resource limited smart devices and also encourage the APs to contribute resource in local model update verification, we design an iterative double auction based joint cooperative federated learning and local model update verification scheme. The optimal unified time for cooperative federated learning and local model update verification is solved to maximize social welfare. Numerical results illustrate that the proposed scheme is efficient in digital twin construction.

With the increasing application of power monitoring terminal and the increasing demand for its function, it is necessary to upgrade the field embedded software remotely.The existing online upgrade schemes generally deal with the terminals one by one, and there are also distributed upgrade schemes. However, due to the large differences in the communication performance of terminals in different places, there is a problem of load imbalance, resulting in the low efficiency of the whole upgrade. Aiming at the efficiency of remote online upgrade of power monitoring terminal, a distributed load balancing remote upgrade method is proposed. The power monitoring terminal is upgraded at the same time through the cloud platform, and the genetic algorithm is used for load balancing processing, so as to improve the efficiency. The practical results show that this remote upgrade method improves the upgrade efficiency of the terminal by comparing a variety of upgrade schemes, compared with no load balancing method, the average time of upgrading a terminal is reduced by 36%, which achieves good application effect in practical application.

A united-reference fractional counter (URFC) is proposed for multi-level time-to-digital converters (TDCs) used in direct time-of-flight (dToF) light detection and ranging (LiDAR) applications. The URFC combines the integer counter and the second-level interpolation in one functional block that derives both the integer part and the fractional part from the same synchronization result. Thus the proposed TDC has a more robust structure to deal with misalignment errors, and measurement results from different levels can be perfectly matched. High linearity and high precision can be guaranteed because of the united reference without calibration. The response time of the URFC decreases to hundreds of picoseconds, which makes it suitable for all output pulses from the avalanche photodiode. A TDC with the proposed URFC was implemented in 65nm standard CMOS technology with a detection range of 2.5μs (~375m) and a single shot precision of 27.63ps (~4.0mm). The power consumption is 51.4mW with a reference clock at 200MHz and a repetition rate at 999kHz. The INL and DNL are ±2.751 LSB and ±1.679 LSB, respectively.

In this paper, we propose a waveform design based electronic counter-countermeasures (ECCM) scheme against interrupted sampling repeater jamming (ISRJ) for wideband multiple-input multiple-output (MIMO) radar. The ECCM scheme includes two steps: ISRJ parameters estimation and ISRJ signal suppression. The ISRJ signal interpretations based on time-frequency (TF) domain and Doppler domain are proposed respectively. Based on the Doppler interpretation, a waveform with thumbtack ambiguity function and a corresponding multiple Doppler channels processing structure are proposed to obtain ISRJ parameters. According to the TF interpretation and the known ISRJ parameters, a waveform with desired TF distribution and a corresponding frequency filter processing strategy are proposed to suppress ISRJ signal. The proposed ECCM scheme can be directly used in phased-array mode without waveform optimization. Considering the multiple targets scenario, the simultaneous multiple beams technology in MIMO radar is preferred. Therefore, a constant modulus waveform design method is proposed to match the desired transmit beampattern and TF distribution. Besides, in order to further suppress the sidelobe output of ISRJ signal, an extended version of the proposed ECCM scheme in the slow-time dimension is proposed. Finally, the numerical simulations are provided to demonstrate the effectiveness of the proposed ECCM scheme in multiple targets scenario.

There are two principle approaches for learning in artificial intelligence: error-driven global learning and neuroscience-oriented local learning. Integrating them into one network may provide complementary learning capabilities for versatile learning scenarios. At the same time, neuromorphic computing holds great promise, but still needs plenty of useful algorithms and algorithm-hardware co-designs to fully exploit its advantages. Here, we present a neuromorphic global-local synergic learning model by introducing a brain-inspired meta-learning paradigm and a differentiable spiking model incorporating neuronal dynamics and synaptic plasticity. It can meta-learn local plasticity and receive top-down supervision information for multiscale learning. We demonstrate the advantages of this model in multiple different tasks, including few-shot learning, continual learning, and fault-tolerance learning in neuromorphic vision sensors. It achieves significantly higher performance than single-learning methods. We further implement the model in the Tianjic neuromorphic platform by exploiting algorithm-hardware co-designs and prove that the model can fully utilize neuromorphic many-core architecture to develop hybrid computation paradigm., Global and local learning represent two distinct approaches to artificial intelligence. In this manuscript, Wu et al present a hybrid learning strategy, drawing from elements of both approaches, and implement it on a co-designed neuromorphic platform.

In this paper, we uniquely combine power-gating and DVFS with the aim of maximizing the NoC power savings and improving performance. The proposed NoC design, called Agile, consists of several architectural designs and a reinforcement learning (RL) based control policy to mitigate the negative effects induced by the combined power-gating and DVFS. Specifically, a simple bypass switch is deployed to maintain network connectivity, avoiding frequently waking up the powered-off router. An optimized pipeline can simplify pipeline stages of the bypass switch to reduce network latency. Reversible link channel buffers can be dynamically allocated to where they are needed to improve throughput. In addition, the RL control policy predicts NoC traffic and decides optimal power-gating decisions, voltage/frequency levels and NoC architecture configurations at runtime. Furthermore, we explore the use of an artificial neural network (ANN) to efficiently reduce the area overhead of implementing RL. We evaluate our design using the PARSEC benchmarks suite. The full system simulation results show that the proposed design improves the overall power savings by up to 58% while improving the performance up to 11% as compared to state-of-the-art designs.

This paper presents the overall comparison of four possible topologies of linear switched reluctance motor with segmented secondary (LSRMSS) as candidates for vertical actuation systems. The analytical design approach which treats each unit of LSRM as independent force generator is applied to design the four machines. All the four machines can be seen as 2-unit force generator although they have different winding configurations and topologies. In this case, the four machines can be designed to have similar force profile due to the inherent modular characteristics of LSRMSS. Force production, active material mass, end effects, copper loss and thrust density have been considered as well as the dynamic performance. The results indicate that the topology with toroidally wound mover has the highest force density and payload ratio capability. This topology is chosen to be prototyped for a 1m long vertical propulsion actuator. The experiment is performed to validate the theoretical analysis above.

Purpose To find the system with minimum investment and best quality performance that is capable of producing all of the product variants, assessing the complexity of designing assembly system at the early concept stage is an essential step, which helps and instructs a designer to create a product- and system-oriented assembly solution with the least complexity. The purpose of this paper is to propose a quantifying measurement of complexity in the design of a modular automated assembly system. Design/methodology/approach The configurable assembly system is becoming a trend, which enables companies to quickly respond to changes caused by different product variants but without a large investment. One of the enabling factors is the availability of modular solutions of assembly modules that can be configured according to different technical requirements. This paper develops a methodology using fuzzy evaluation to calculate the design complexity in the design phase for a modular automatic assembly system. Fuzzy linguistic variables are used to measure the interaction among the influence factors, to deal with the uncertainty of the judgement. The proposed method investigates three matrices to present how the function-based assembly modules, design complexity factors, part attributes and product components, which are regarded as the main influence factors, complicate the construction of a modular assembly system. The design complexity is derived and quantified based on these assessments. Findings The proposed approach presents a formal quantification to evaluate the design complexity with regard to a modular assembly system from beginning, which can be identified and used as criteria to indicate the quality of performance and investment cost in advance. A mathematical model based on the fuzzy logic is established to provide both theoretical and practical guidance for the paper. To validate the predictive model, the statistic relationships between the assessed system design complexity, real assembly defect rate and investment cost are estimated based on regression analysis. The application of the presented methodology is demonstrated with regard to a traditional rear drive unit in the automotive industry. Originality/value This paper presents a developed method, which addresses the measures of complexity found in the design of a modular assembly system. It would help to run the design process with better resource allocation and cost estimation in a quantitative approach.

Object occlusion is a common issue in Light Detection and Ranging (LiDAR)-based vehicle tracking technology. The occlusions can cause variance in vehicle location and speed calculation. How to link the vehicle trajectories caused by occlusion issues is a challenge for traffic engineers and researchers. This paper developed an augmented vehicle tracking method under occlusions with the roadside LiDAR data. The proposed method can be divided into two parts. The first part based on the corner point is used to choose a representative vehicle tracking point. And the second part based on the GNN algorithm is employed to link the vehicles’ trajectories under two occlusion situations. The performance of the proposed method has been evaluated using roadside lidar data collected from four different scenarios. The test results showed that more than 89% of disconnected trajectories can be fixed with the proposed method, which is superior compared to the state-of-the-art method. The proposed method can benefit a lot of transportation areas, such as traffic volume count, vehicle speed tracking, and traffic safety analysis.

Non‐uniform irradiance, for example, partial shading, would affect the performance of the whole Photovoltaic (PV) array and reduce the output power. Since the layout of PV array is also an important factor influencing the performance of PV array, reconfiguration scheme has been researched popularly. Since plenty of array reconfiguration techniques are not realistic to be implemented in reality, in this paper considering the simplicity, robustness and economical aspects, a Followed The Regularized Leader algorithm (FTRL) regression prediction model based switches controlled PV array reconfiguration method is proposed to enhance the power output from solar PV array and alleviate the effects of partial shadings. The proposed scheme has no additional labor and wire requirement, since the proposed connection architecture of PV array involves no complex switch matrix design. Subsequently, to testify the performance of the proposed scheme, simulations with various shades in different PV arrays are made and evaluation parameters, for example, Fill Factor (FF) and mismatch loss are calculated. Furthermore, outdoor experiments are conducted to validate the simulation and the analyses. Besides, extensive analyses of cost analysis and energy savings are given. At last, a comprehensive evaluation is performed to check the quality of the proposed scheme by comparing with various techniques in literature.

Virtual synchronous generator technology can effectively improve the anti‐interference characteristics of the system frequency and bus voltage in the microgrid, and solve the problems of insufficient damping and low inertia of the system. However, in an islanded microgrid with multiple distributed generation, the difference in line impedance will cause local voltage deviations, which in turn leads to a series of problems, such as reducing power distribution accuracy and increasing bus voltage drop. Therefore, for the island‐type microgrid multi‐inverter distributed power generation parallel system, in order to solve the problem of low power distribution accuracy and large frequency oscillation caused by system parameters in virtual synchronous generator control, an improved virtual synchronous generator control algorithm based on adaptive droop coefficient is proposed in this paper, which not only eliminates the resistance component of the line impedance, makes the system impedance characteristic present a purely inductive nature, but also realises real‐time adjustment of active and reactive power. While maintaining the stability of the bus voltage and system frequency, it maintains high power distribution accuracy and improves the dynamic performance and operational stability of the power grid system.

Steerable drills have the potential to minimize the required dimensions of incisions in minimally invasive spine fusion (MISF). Existing steerable drills mostly rely on flexible continuum mechanisms to create distal tip steerability. However, due to their inherent compliance and underactuation, high stiffness and dexterity cannot be achieved at the same time, which limits their application. In this article, we present a omm robotic steerable surgical drill offering high stiffness, strength, and dexterity, simultaneously. Contrary to existing flexible-transmission-based approaches, we use nonflexible transmissions (tendon-driven articulated joints and universal joints) to provide two distal degrees of freedom. To evaluate its performance, we integrated it with a customized tendon actuation platform for mechanical performance tests and the da Vinci Research Kit, as an exchangeable instrument, for robotic steerable drilling tests. The results of these tests showed that our drill had high strength, stiffness, accuracy, and drilling stability while preserving high steerability and was sufficient to complete hole drilling in confined space in MISF.

Selective disassembly is a systematic method to remove target components or high-valuable components from an EOL product for reuse, recycling and remanufacturing as quick and feasible as possible, which plays a key role for the effective application of circular economy. However, in practice, the process of selective disassembly is usually characterized by various unpredictable factors of EOL products. It is very difficult to identify a feasible disassembly sequence for getting the target components before taking actions due to the uncertainty. In this paper, a data-driven method of selective disassembly planning for EOL products under uncertainty is proposed, in which disassemblability is regarded as the degree of difficulty in removing components under uncertainty. Taxonomy of uncertainty metrics that represents uncertain characteristics of components and disassembly transitions of selective disassembly is established. Random and fuzzy assessment data of uncertainty is converted into qualitative values and aggregated to fit a prediction model based on the trapezium cloud model. The turning time of disassemblability is predicted for a given set of certainty degree. Further, the disassemblability values are applied to determine the best selective disassembly sequence in order to get target component with tradeoff between minimum number of disassembly operations and maximum feasibility. The effectiveness of the proposed method is illustrated by a numerical example. Moreover, by comparing to selective disassembly planning without considering uncertainty, the proposed method turns selective disassembly of EOL products more realistic than 11% and provide insights on how to design product to facilitate disassembly operations.

Four-dimensional (4D) printing is an advanced form of three-dimensional (3D) printing with controllable and programmable shape transformation over time. Actuators are used as a controlling factor with multi-stage shape recovery, with emerging opportunities to customize the mechanical properties of bio-inspired structures. The print pattern of shape memory polymer (SMP) fibers strongly affects the achievable resolution, and consequently influences several other physical and mechanical properties of fabricated actuators. However, the deformations of bio-inspired structures due to actuator layout are more complex because of the presence of the coupling of multi-directional strain. In this study, the initial structure was designed from closed-shell behavior and divided into a general unit and actuator unit, the latter responsible for driving the transformation. Mutual stress confrontation between the actuator and the general unit was considered in the layout thermodynamic model, in order to eliminate the transformation produced by the uncontrolled shape memory behavior of the general unit. Three critical and effective strategies for the layout design of actuators were proposed and then applied to achieve the desired accurate deformation of 3D-printed bilayer structures. Finally, the proposed approach was validated and adopted for fabricating a complex shell-like gripper structure.

With the development of city size and vehicle interconnection, visual analysis technology is playing a very important role in the course of city calculation and city perception. A Reasonable visual model can effectively present the feature of city. In order to solve the problem of traditional density algorithm that cluster the large scale data slowly and cannot find cluster centers to adapt taxi track data. The DBSCAN+ (density-based spatial clustering of applications with noise plus) algorithm that can split data and extract maximum density clusters under the large scale data was proposed in the paper. The passenger points should be cleaned from the original point of the passenger trajectory data firstly, and then the massive passenger points are sliced and clustered cyclically. In the clustering process, the cluster centers can be extracted based on maximum density, and finally the clustering results are visualized according to the results. The experimental results show that compared with other popular methods, the proposed method has significant advantages in clustering speed, precision and visualization for large-scale city passenger hotspots. Moreover, it provides important decisions for further urban planning and promotes the traffic efficiency.

Many classical feature extraction and dimensionality reduction algorithms, such as linear algorithm principal component analysis (PCA) and nonlinear algorithm local linear embedding (LLE), have been applied to face recognition. As we all know, PCA is a global algorithm and LLE is a local algorithm. However, PCA cannot obtain the local structure of high-dimensional spatial data samples, and LLE cannot obtain the global structure of high-dimensional spatial data samples. The application effect of these algorithms is not ideal, especially when they are always affected by overlapping points (outliers) and sparse points in the data. To solve these problems, the paper proposes a new effective feature extraction and dimension reduction algorithm called sparse fuzzy difference embedding (SFDE). Firstly, SFDE algorithm tries to search an optimal projection mapping matrix, which can affect not only the local of the fuzzy local minimizing embedding obtained by LLE but also the global of the fuzzy global maximizing variance obtained by PCA. Secondly, SFDE algorithm obtains the sparse transformation matrix by using the lasso regression return. This feature makes SFDE more intuitive and powerful than PCA, LLE, and other algorithms. Finally, we estimated the proposed algorithm through experiments in Yale and AR standard face databases and added the density of “salt and pepper” noise to the Yale and AR databases to verify the robustness of SFDE algorithm. The experimental results of the SFDE algorithm were better than those of the LDA, PCA, LLE, sparse differential embedding (SDE), and fuzzy local graph embedding based on maximum margin criterion algorithms because of its sparsity and fuzzy set, which also indicates that the SFDE algorithm is an effective algorithm.

We present V2V, a novel deep learning framework, as a general-purpose solution to the variable-to-variable (V2V) selection and translation problem for multivariate time-varying data (MTVD) analysis and visualization. V2V leverages a representation learning algorithm to identify transferable variables and utilizes Kullback-Leibler divergence to determine the source and target variables. It then uses a generative adversarial network (GAN) to learn the mapping from the source variable to the target variable via the adversarial, volumetric, and feature losses. V2V takes the pairs of time steps of the source and target variable as input for training, Once trained, it can infer unseen time steps of the target variable given the corresponding time steps of the source variable. Several multivariate time-varying data sets of different characteristics are used to demonstrate the effectiveness of V2V, both quantitatively and qualitatively. We compare V2V against histogram matching and two other deep learning solutions (Pix2Pix and CycleGAN).

Successful product development relies on the enterprise system (ES) to manage product lifecycle data and support decision-making in various levels. Since the amount of data generated in the industrial Internet of Things is increasing dramatically, new paradigms of the ES are required to realize the distributed intelligence. Recent progress in edge computing has enabled advances in decentralized decision support systems. Although the product design is a key stage of new product development, none of the recent studies in edge-based ES have shed light on this stage. Therefore, an edge-based ES framework for design scheme evaluation is proposed in this article. It not only accomplishes most of the decision-making tasks within the edge mesh of an enterprise, but also solves the information island of multiple decision makers from different geographical regions. Moreover, a multigroup decision-making algorithm is proposed to enable collaborative design scheme evaluation. The evaluation processes of designers, experts, and customers are analyzed systematically in this article. For one advantage, the trapezium cloud model is applied to convert the qualitative evaluation information given by designers and experts into qualitative values. It decreases the cognitive discrepancy and solves the information distortion. For another, EEG data are utilized to explore the implicit psychological states of customers during the product operation. All the evaluation results of multiple groups of decision makers are integrated by the fuzzy measure and Choquet integral to determine the optimal design scheme. A case study is conducted to illustrate the feasibility of the method proposed in this article.

Compared with static shape control, dynamic shape control is more difficult due to its time-varying nature. To improve the control precision and reduce the number of actuators needed, an integrated design optimization model of structure and control is proposed. The following three sets of design variables are considered to minimize the time domain variance between the controlled and desired dynamic shapes: the time-varying actuator voltages, the actuator layout and the structural ply parameters. To satisfy the engineering performance requirements to the greatest extent possible, constraints on the structural mass, the energy, the maximum transient voltage and the number of actuators are considered. Because the control voltages are varying in time, a two-level optimization strategy is adopted. In the inner optimization problem, the Newmark integral method is applied to derive discrete time domain expressions for the shape control equations, and then, the Kuhn-Tucker condition is introduced to calculate the optimal time-varying voltage distribution. To solve the outer optimization problem, because of the coexistence of discrete variables and continuous variables, a simulated annealing algorithm is used. To address the shape control of complicated curved shells, a finite element formulation for an eight-node laminated curved shell element with piezoelectric actuators is derived. Numerical examples show that the proposed integrated design optimization method can significantly improve the control effect and that the optimization of the structural ply parameters plays an important role. Moreover, the control system can be simplified by taking the minimum number of actuators as the objective function when the control accuracy allows.

Predictive maintenance is one of the important technical means to guarantee and improve the normal industrial production. The existing bottlenecks for popularization and application are analyzed. In order to solve these problems, a cooperative awareness and interconnection framework across multiple organizations for total factors that affect prediction maintenance decision-making is discussed. Initially, the structure and operation mechanism of this framework are proposed. It is designed to support the sharing of data, knowledge and resources. As a key supporting technology, the digital twin is also integrated into it to improve the accuracy of fault diagnosis and prediction and support making a maintenance plan with higher accuracy and reliability. Then, under this framework, an integrated mathematical programming model is established with considering the parameter uncertainty and an NSGA-II hybrid algorithm is utilized to solve it. Moreover, an adjustment strategy for a maintenance plan is discussed in response to the dynamic characteristics of the actual maintenance environment. Finally, a case, prediction maintenance decision-making for bearings in grinding rolls of the large vertical mill, is studied. Analysis results verify the advantage of the integrated solving mechanism based on the proposed framework. The framework and integrated decision-making approach can guide the implementation of predictive maintenance with higher accuracy and reliability for industrial enterprises.

In the process of human behavior recognition, the traditional dense optical flow method has too many pixels and too much overhead, which limits the running speed. This paper proposed a method combing YOLOv3 (You Only Look Once v3) and local optical flow method. Based on the dense optical flow method, the optical flow modulus of the area where the human target is detected is calculated to reduce the amount of computation and save the cost in terms of time. And then, a threshold value is set to complete the human behavior identification. Through design algorithm, experimental verification and other steps, the walking, running and falling state of human body in real life indoor sports video was identified. Experimental results show that this algorithm is more advantageous for jogging behavior recognition.

Purpose Volumetric pancreas segmentation can be used in the diagnosis of pancreatic diseases, the research about diabetes and surgical planning. Since manual delineation is time-consuming and laborious, we develop a deep learning-based framework for automatic pancreas segmentation in three dimensional (3D) medical images. Methods A two-stage framework is designed for automatic pancreas delineation. In the localization stage, a Square Root Dice loss is developed to handle the trade-off between sensitivity and specificity. In refinement stage, a novel 2.5D slice interaction network with slice correlation module is proposed to capture the non-local cross-slice information at multiple feature levels. Also a self-supervised learning-based pre-training method, slice shuffle, is designed to encourage the inter-slice communication. To further improve the accuracy and robustness, ensemble learning and a recurrent refinement process are adopted in the segmentation flow. Results The segmentation technique is validated in a public dataset (NIH Pancreas-CT) with 82 abdominal contrast-enhanced 3D CT scans. Fourfold cross-validation is performed to assess the capability and robustness of our method. The dice similarity coefficient, sensitivity, and specificity of our results are 86.21 ± 4.37%, 87.49 ± 6.38% and 85.11 ± 6.49% respectively, which is the state-of-the-art performance in this dataset. Conclusions We proposed an automatic pancreas segmentation framework and validate in an open dataset. It is found that 2.5D network benefits from multi-level slice interaction and suitable self-supervised learning method for pre-training can boost the performance of neural network. This technique could provide new image findings for the routine diagnosis of pancreatic disease.

Improving the energy efficiency of machining systems (MSs) is an urgent and challenging issue in sustainable manufacturing. Considering the dynamics and uncertainty of MSs, a five-dimension model is developed to monitor and reduce the energy consumption (EC) of MSs. In this model, the automatic acquisition of EC data from physical MSs is driven by Radio Frequency IDentification (RFID) read events. The EC service model provides monitoring and scheduling services for reducing the EC of physical MSs by connecting with virtual MSs in real time. The procedures of the EC reduction method for the MSs are presented. A case study of MSs for auto parts was conducted to illustrate the proposed method. The results showed that the time and EC of MSs were reduced by 148 s and 49.67 KJ, respectively.

As a state-of-the-art multi-class supervised novelty detection method, supervised novelty detection-support vector machine (SND-SVM) is extended from one-class support vector machine (OC-SVM). It still requires to slove a more time-consuming quadratic programming (QP) whose scale is the number of training samples multiplied by the number of normal classes. In order to speed up SND-SVM learning, we propose a down sampling framework for SND-SVM. First, the learning result of SND-SVM is only decided by minor samples that have non-zero Lagrange multipliers. We point out that the potential samples with non-zero Lagrange multipliers are located in the boundary regions of each class. Second, the samples located in boundary regions can be found by a boundary detector. Therefore, any boundary detector can be incorporated into the proposed down sampling framework for SND-SVM. In this paper, we use a classical boundary detector, local outlier factor (LOF), to illustrate the effective of our down sampling framework for SND-SVM. The experiments, conducted on several benchmark datasets and synthetic datasets, show that it becomes much faster to train SND-SVM after down sampling.

Deep multi-task learning (DMTL) is an efficient machine learning technique that has been widely utilized for facial expression recognition. However, current deep multi-task learning methods typically only consider the information of class labels, while ignoring the local information of sample spatial distribution. In this paper, we propose a discriminative DMTL (DDMTL) facial expression recognition method, which overcomes the above shortcomings by considering both the class label information and the samples’ local spatial distribution information simultaneously. We further design a siamese network to evaluate the local spatial distribution through an adaptive reweighting module, utilizing the class label information with different confidences. In addition, by taking the advantage of the provided local distribution information of samples, DDMTL is able to achieve acceptable results even if the number of training samples is small. We implement experiments on three facial expression datasets. The experimental results demonstrate that DDMTL is superior to the state-of-the-art methods.

Networks-on-chip (NoCs) are playing a critical role in modern multicore architecture, and NoC security has become a major concern. Maliciously implanted hardware Trojans (HTs) inject faults into on-chip communications that saturate the network, resulting in the leakage of sensitive data via side channels and significant performance degradation. While existing techniques protect NoCs by detecting and isolating HT-infected components, they inevitably incur occasional inaccurate detection with considerable network latency and power overheads. We propose TSA-NoC, a learning-based design framework for secure and efficient on-chip communication. The proposed TSA-NoC uses an artificial neural network for runtime HT-detection with higher accuracy. Furthermore, we propose a deep-reinforcement-learning-based adaptive routing design for HT mitigation with the aim of minimizing network latency and maximizing energy efficiency. Simulation results show that TSA-NoC achieves up to 97% HT-detection accuracy, 70% improved energy efficiency, and 29% reduced network latency as compared to state-of-the-art HT-mitigation techniques.

Unsupervised domain adaptation, which aims to classify a target domain correctly only using a labeled source domain, has achieved promising performance yet remains a challenging problem. Most traditional methods focus on exploiting either geometric or statistical characteristics to reduce domain shifts. To take advantage of both sides, in this paper, we propose a unified framework incorporating both the geometric and statistical characteristics by adopting the non-convex Schatten p-norm and graph Laplacian constraints to preserve global and local structure information and constructing marginal and conditional distribution minimization terms to reduce the distribution shifts. Moreover, a classification error term on the source domain is embedded into the objective function to increase the discriminability. The proposed method has been evaluated on six datasets and the experimental results demonstrate the superiority of the proposed method over several state-of-the-art methods. The MATLAB code of our method will be publicly available at https://github.com/HeyouChang/unsupervised-domain-adaptation .

TV- $$L^1$$ is a classical diffusion–reaction model for low-level vision tasks, which can be solved by a duality-based iterative algorithm. Considering the recent success of end-to-end learned representations, we propose a TV-LSTM network to unfold the duality-based iterations of TV- $$L^1$$ into long short-term memory (LSTM) cells. In particular, we formulate the iterations as customized layers of a LSTM neural network. Then, the proposed end-to-end trainable TV-LSTMs can be naturally connected with various task-specific networks, e.g., optical flow, image decomposition and event-based optical flow estimation. Extensive experiments on optical flow estimation and structure + texture decomposition have demonstrated the effectiveness and efficiency of the proposed method.

As a representative method of model surface, triangular mesh, which commonly has the stereo lithography (STL) format, has recently been widely applied in the CAD/CAE/CAM field, due to its superior robustness and high efficiency in tool-path generation. 3D product model optimization is of great importance for improvement of function, reduction of production material, and improvement of the company’s competition. Based on the circumstance of 3D modeling, simulation and optimization technologies, developing a more sustainable product and process become possible. However, there are many situations of the morphing, which are hard to be uniformed. Thus, developing a commonly used data-driven morphing method is difficult. In this paper, morphing situations are categorized into two classes, the algebraic morphing and the free-form morphing. Algebraic morphing patterns are developed, which can be adopted independently or combined together to complete complicated morphing operations. In the free-form morphing, control points are obtained by data mining, and then mesh subdivision is applied to refine surfaces smoothly. The proposed morphing method is applied to a truss core panel and a human head model, clarifying the robust function and high efficiency of the method proposed in this study to deal with complex 3D product model sustainable optimization.

MotivationThe generalized linear mixed model (GLMM) is an extension of the generalized linear model (GLM) in which the linear predictor takes random effects into account. Given its power of precisely modeling the mixed effects from multiple sources of random variations, the method has been widely used in biomedical computation, for instance in the genome-wide association studies (GWASs) that aim to detect genetic variance significantly associated with phenotypes such as human diseases. Collaborative GWAS on large cohorts of patients across multiple institutions is often impeded by the privacy concerns of sharing personal genomic and other health data. To address such concerns, we present in this paper a privacy-preserving Expectation–Maximization (EM) algorithm to build GLMM collaboratively when input data are distributed to multiple participating parties and cannot be transferred to a central server. We assume that the data are horizontally partitioned among participating parties: i.e. each party holds a subset of records (including observational values of fixed effect variables and their corresponding outcome), and for all records, the outcome is regulated by the same set of known fixed effects and random effects.ResultsOur collaborative EM algorithm is mathematically equivalent to the original EM algorithm commonly used in GLMM construction. The algorithm also runs efficiently when tested on simulated and real human genomic data, and thus can be practically used for privacy-preserving GLMM construction. We implemented the algorithm for collaborative GLMM (cGLMM) construction in R. The data communication was implemented using the rsocket package.Availability and implementationThe software is released in open source at https://github.com/huthvincent/cGLMM.Supplementary informationSupplementary data are available at Bioinformatics online.

Many studies have approved that optimising the cutting parameters is effective for reducing the energy consumption of machining operations of machine tools. However, this technique to reduce the end face turning energy consumption (EFTEC), where the material removal rate is variable, has not received attention. Besides, the energy consumed for the spindle speed changes was ignored in previous research. Aiming to fill these gaps, an integrated EFTEC model is developed considering the spindle speed changes. In terms of optimisation, the EFTEC model is discretised according to the allowable accuracy of the machine tool. Simulated annealing is adopted to search for the optimal values of cutting parameters that lead to the minimum EFTEC. In the case study, nine parts with changing diameters and cutting depths are machined by a lathe (CK6153i). According to the experiments, simulated annealing has more than 96% probability of obtaining the global optima. The optimum achieves a 14.03% EFTEC reduction for a case. The relationship between the design parameters and the optimal cutting parameters is discussed. A case shows that 2.43% of the machining time increases suffer from the EFTEC optimisation.

We present AntVis, a web-based visual analytics tool for exploring ant movement data collected from the video recording of ants moving on tree branches. Our goal is to enable domain experts to visually explore massive ant movement data and gain valuable insights via effective visualization, filtering, and comparison. This is achieved through a deep learning framework for automatic detection, segmentation, and labeling of ants, ant movement clustering based on their trace similarity, and the design and development of five coordinated views (the movement, similarity, timeline, statistical, and attribute views) for user interaction and exploration. We demonstrate the effectiveness of AntVis with several case studies developed in close collaboration with domain experts. Finally, we report the expert evaluation conducted by an entomologist and point out future directions of this study. Keywords: Ant movement, Object detection, Image segmentation, Visual analytics, Knowledge discovery

In this paper, resource flow variables are extracted from the internal structural features of the logistics distribution process and a new method for optimizing the internal structure of the logistics distribution system by using the characteristic state space is proposed. The characteristic state equation is constructed to represent the input and output resources of each basic logistics activity. The basic logistics activity equation is iterated according to the resource flow, and the implementation of the basic logistics process is visually and quantitatively expressed in the form of the characteristic state matrix. According to the nature of the characteristic state space, the optimization problem of the logistics distribution system is transformed into a critical path-planning problem, the gradient calculation of the objective function is solved, and an improved genetic algorithm is proposed. This accelerates the convergence speed of the algorithm and reduces the running time of the optimization process. Taking a listed logistics distribution enterprise as an example, the optimization algorithm is verified, which proves the advantages of the algorithm and provides a new method and theoretical basis for the analysis and optimization of the logistics distribution system.

In an islanded microgrid with multiple distributed generations (DGs), the difference in line impedance may cause local voltage deviation, which leads to a series of problems such as lower power allocation accuracy and bus voltage drop under traditional droop control. In this respect, a method for optimizing the droop control using an improved particle swarm optimization (IPSO) is proposed. Firstly, the microgrid structure and influence of line parameters on traditional droop control strategy is analyzed. Then, an improved particle swarm optimization is proposed. Based on the basic particle swarm optimization (PSO) algorithm, a fuzzy inference system (FIS) is introduced to dynamically adjust the particle swarm optimization, which can effectively improve the global search ability and local search ability of the algorithm. After that, the improved algorithm is applied to the droop controller, simultaneously, the range of stable operation of the system is determined by small signal analysis. Finally, the simulation and experiment results show that the proposed improved droop control strategy can achieve accurate allocation of active and reactive power effectively while maintaining bus voltage and system frequency stability, enhance the dynamic performance and transient stability of the microgrid system.

As an important link to guarantee normal industrial production, equipment maintenance plays an increasingly key role in enhancing the competitiveness of enterprises and supporting green smart manufacturing. This paper aims to promote the implementation of predictive maintenance for complex equipment and improve the green performance of the maintenance service process. A structural framework of information sharing and service network is introduced to build a ubiquitous state data awareness environment for predictive maintenance service. Subsequently, an integrated mathematical problem model that consists of carbon emission objective and extended maintenance cost objective is constructed. Then an improved NSGA-II algorithm is utilized to solve this complicated two-objective optimization problem. In response to deal with the uncertainties of maintenance service environment and inaccuracy of prediction, a data-driven dynamic adjustment strategy is applied. A grinding roll fault case of a large vertical is used to demonstrate the effectiveness of this proposed approach.

Although the current vehicle detection and recognition framework based on deep learning has its own characteristics and advantages, it is difficult to effectively combine multi-scale and multi category vehicle features, and there is still room for improvement in vehicle detection and recognition performance. Based on this, an improved fast R-CNN convolutional neural network is proposed to detect dim targets in complex traffic environment. The deep learning model of fast R-CNN convolutional neural network is introduced into the image recognition of complex traffic environment, and a structure optimization method is proposed, which replaces vgg16 in fast RCNN with RESNET to make it suitable for small target recognition in complex background. Max pooling is the down sampling method, and then feature pyramid network is introduced into RPN to generate target candidate box to optimize the structure of convolutional neural network. After training with 1497 images, the complex traffic environment images are identified and tested.

Incast occurs frequently in datacenter networks where a large number of senders send data to a single receiver simultaneously, which makes the last hop the network bottleneck. Incast can hurt flows' performance. However, congestion control protocols are not effective at handling incast. One key insight is that it is too late to handle incast packets after they have already piled up at the last hop. Instead, we should avoid incast as early as possible. Inspired by flood control in Hydrologic Engineering, we propose Floodgate, a novel switch-based per-hop flow control to handle incast. Floodgate is compatible with existing congestion control protocols. We integrate it with practical congestion control approaches such as DCQCN, TIMELY, and HPCC. We evaluate Floodgate both in our implementations and large-scale simulations. Compared with state of the art, Floodgate reduces the buffer occupancy by a factor of 6.6x, as well as the queuing delay. Therefore, the average FCT and tail latency are greatly reduced.

This paper presents a comparative analysis of two parallel hybrid magnet memory machines (PHMMMs) with different permanent magnet (PM) arrangements. The proposed machines are both geometrically characterized by a parallel U-shaped hybrid PM configuration and several q-axis magnetic barriers. The configurations and operating principles of the investigated machines are introduced firstly. The effect of magnet arrangements on the performance of the proposed machines is then evaluated with a simplified magnetic circuit model. Furthermore, the electromagnetic characteristics of the proposed machines are investigated and compared by the finite-element method (FEM). The experiments on one prototype are carried out to validate the FEM results.

Technological advancements have led to easy and rapid generation and collection of huge amounts of varieties of data from of wide ranges of rich data sources. These big data may be of different levels of veracity, including precise data and imprecise or uncertain data. Embedded in the data are valuable information and useful knowledge that can be discovered by data analytics. Discovered information and knowledge may help to build a smart city and then a smart world. In this paper, we focus on making good fusion of conceptual modelling and knowledge graphs to capture essential data about public transportation (e.g., buses). Specifically, our conceptual model and knowledge graph capture information regarding bus arrival and departure. Some of the captured data can be uncertain (e.g., timestamp, GPS locations of buses) due to the limitations of the measuring devices and methods to collect the data. Nonetheless, the models are helpful in smart city big data analytics of these public transportation data. The discovered knowledge produces insights to users (e.g., city planners, policy makers), which in turn help them to build a smart city.

Technological advancements have led to easy and rapid generation and collection of huge volumes of varieties of data from of wide ranges of rich data sources. These big data may be of different levels of veracity, including precise data and imprecise or uncertain data. Embedded in the data are valuable information and useful knowledge that can be discovered by data mining. Discovered information and knowledge may help to build a smart world. In this paper, we present a solution for scalable mining of big data. In particular, we focus on scalable mining of huge volumes of temporal coronavirus disease 2019 (COVID-19) data at different granularity levels. Since its outbreak, there have been millions of COVID-19 cases worldwide. These are huge volumes of data, and new cases have been reported every day. Embedded in these COVID-19 data is implicit, previously unknown and potentially useful information and knowledge, which can be discovered by data mining for social good. Analyzing and mining these data helps users (e.g., researchers, civilian) to get better understanding of the disease, and thus take an active role in fighting, controlling, and/or combating the disease. Evaluation results on real-life COVID-19 data show the benefits of our solution in scalable mining of big data.

We present STNet, an end-to-end generative framework that synthesizes spatiotemporal super-resolution volumes with high fidelity for time-varying data. STNet includes two modules: a generator and a spatiotemporal discriminator. The input to the generator is two low-resolution volumes at both ends, and the output is the intermediate and the two-ending spatiotemporal super-resolution volumes. The spatiotemporal discriminator, leveraging convolutional long short-term memory, accepts a spatiotemporal super-resolution sequence as input and predicts a conditional score for each volume based on its spatial (the volume itself) and temporal (the previous volumes) information. We propose an unsupervised pre-training stage using cycle loss to improve the generalization of STNet. Once trained, STNet can generate spatiotemporal super-resolution volumes from low-resolution ones, offering scientists an option to save data storage (i.e., sparsely sampling the simulation output in both spatial and temporal dimensions). We compare STNet with the baseline bicubic+linear interpolation, two deep learning solutions ( SSR+TSF, STD), and a state-of-the-art tensor compression solution (TTHRESH) to show the effectiveness of STNet.

Health environment is a key factor in public health. Since people’s health depends largely on their lifestyle, the built environment which supports a healthy living style is becoming more important. With the right urban planning decisions, it’s possible to encourage healthier living and save healthcare expenditures for the society. However, there is not yet a quantitative relationship established between urban planning decisions and the health status of the residents. With the abundance of data and computing resources, this research aims to explore this relationship with a machine learning method. The data source is from both the OpenStreetMap and American Center for Decease Control and Prevention (CDC). By modeling the Point of Interest data and the geographic distribution of health-related outcome, the research explores the key factors in urban planning that could influence the health status of the residents quantitatively. It informs how to create a built environment that supports health and opens up possibilities for other data-driven methods in this field.

Knee cartilage and bone segmentation is critical for physicians to analyze and diagnose articular damage and knee osteoarthritis (OA). Deep learning (DL) methods for medical image segmentation have largely outperformed traditional methods, but they often need large amounts of annotated data for model training, which is very costly and time-consuming for medical experts, especially on 3D images. In this paper, we report a new knee cartilage and bone segmentation framework, KCB-Net, for 3D MR images based on sparse annotation. KCB-Net selects a small subset of slices from 3D images for annotation, and seeks to bridge the performance gap between sparse annotation and full annotation. Specifically, it first identifies a subset of the most effective and representative slices with an unsupervised scheme; it then trains an ensemble model using the annotated slices; next, it self-trains the model using 3D images containing pseudo-labels generated by the ensemble method and improved by a bi-directional hierarchical earth mover’s distance (bi-HEMD) algorithm; finally, it fine-tunes the segmentation results using the primal-dual Internal Point Method (IPM). Experiments on two 3D MR knee joint datasets (the Iowa dataset and iMorphics dataset) show that our new framework outperforms state-of-the-art methods on full annotation, and yields high quality results even for annotation ratios as low as 5%.

In this paper, we investigate the joint design of a transmit beampattern and angular waveform (AW) for colocated multiple-input multiple-output (MIMO) radars. The importance of the AW in the proposed signal processing strategy is first clarified, and then, two optimization models are established, which are aimed at either the power spectral density (PSD) design or the spectral compatibility and similarity design of the AW. There are two main differences between the proposed models and existing models. First, instead of matching a desired template or maximizing the transmit power on specific regions, the transmit beampattern in this paper is optimized to approach several key points, which guarantees the high transmit gain and the flexible adjustment of each beam gain. Second, instead of optimizing the performance of the transmit waveform, only the characteristics of the AW are examined, and they can be constrained quantitatively according to their relationship with the transmit gain. The two models can be unified into the same framework, and an efficient algorithm is proposed to solve the problem under a constant modulus constraint. The convergence of the proposed algorithm is demonstrated, and some improvements to reduce the computational complexity are proposed. Numerical simulations showed that compared to the existing methods, the proposed approach can be used to obtain a higher transmit gain, flexibly adjust each beam gain, and more accurately control the PSD, spectral compatibility, and similarity of the AW. Moreover, numerical simulations showed that, compared to the use of existing methods, the proposed algorithm has higher computational efficiency.

The current studies of ciphertext watermarking for remote sensing image are mostly focused on the copyright protection. However, the ciphertext data of remote sensing image is at risk of being destroyed, how to ensure the integrity of it is also a problem to be solved. Therefore, aiming at the problem of integrity verification of remote sensing image ciphertext, this paper proposes a ciphertext domain fragile watermarking algorithm by using the multi-permutation superposition coding and the probabilistic encryption. According to the experimental results, the proposed algorithm is highly sensitive to the tamper of ciphertext and can locate the tamper accurately. In addition, the embedding capacity of the proposed ciphertext watermarking is large enough for the copyright protection or other applications, meanwhile, the proposed algorithm will not affect the original remote sensing image date and its future application.

In the current era of big data, huge volumes of valuable data have been generated and collected at a rapid velocity from a wide variety of rich data sources. In recent years, the initiates of open data also led to the willingness of many government, researchers, and organizations to share their data and make them publicly accessible. An example of open big data is healthcare, disease and epidemiological data such as privacy-preserving statistics on patients who suffered from epidemic diseases like the coronavirus disease 2019 (COVID-19). Analyzing these open big data can be for social good. For instance, analyzing and mining the disease statistics helps people to get a better understanding of the disease, which may inspire them to take part in preventing, detecting, controlling and combating the disease. As “a picture is worth a thousand words”, having the pictorial representation further enhances people’s understanding of the data and the corresponding results for the analysis and mining. Hence, in this paper, we present a visual data science solution for the visualization and visual analytics of big sequential data. We illustrate the ideas through the visualization and visual analytics of sequences of real-life COVID-19 epidemiological data. Our solution enables people to visualize COVID-19 epidemiological data and their temporal trends. It also allows people to visually analyze the data and discover relationships among popular features associated with the COVID-19 cases. Evaluation of these real-life sequential COVID-19 epidemiological data demonstrates the effectiveness of our visual data science solution in enhancing user experience in the visualization and visual analytics of big sequential data.

This paper aims to study how existing trace mining methods work, and their strengths and weaknesses in the context of communication-centric system-on-chip (SoC) traces. SoC traces are unique because of the inherent concurrency, which makes SoC trace mining very challenging. We select seven well-known trace mining methods both from the hardware and software domains. We perform comprehensive experiments, and offer our understanding of the pros and cons of each mining method. We evaluate the interestingness of the mined outcomes for each tool on a benchmark trace set. This trace benchmark includes sophisticated communication traces generated synthetically and realistically. We provide a comprehensive analysis of the performance of the tools, and release the benchmark trace dataset to facilitate future works in SoC trace mining.

Purpose Complexity is the main challenge for present and future manufacturers. Assembly complexity heavily affects a product’s final quality in the fully automated assembly system. This paper aims to propose a new method to assess the complexity of modern automated assembly system at the assembly design stage with respect to the characteristics of both manufacturing system and each single component to be mounted. Aiming at validating the predictive model, a regression model is additionally presented to estimate the statistic relationship between the real assembly defect rate and predicted complexity of the fully automated assembly system. Design/methodology/approach The research herein extends the S. N. Samy and H. A. ElMaraghy’s model and seeks to redefine the predictive model using fuzzy evaluation against a fully automated assembly process at the assembly design stages. As the evaluation based on the deterministic scale with accurate crisp number can hardly reflect the uncertainty of the judgement, fuzzy linguistic variables are used to measure the interaction among influence factors. A dependency matrix is proposed to estimate the assembly complexity with respect to the interactions between mechanic design, electric design and process factors and main functions of assembly system. Furthermore, a complexity attributes matrix of single part is presented, to map the relationship between all individual parts to be mounted and three major factors mentioned in the dependency matrix. Findings The new proposed model presents a formal quantification to predict assembly complexity. It clarifies that how the attributes of assembly system and product components complicate the assembly process and in turn influence the manufacturing performance. A center bolt valve in the camshaft of continue variable valve timing is used to demonstrate the application of the developed methodology in this study. Originality/value This paper presents a developed method, which can be used to improve the design solution of assembly concept and optimize the process flow with the least complexity.

Because the fault features of planetary gear are very weak, the feature extraction process of planetary gear fault under traditional thinking is becoming more and more complex, and the amount of extracted features is very small. To solve these problems, a new fault diagnosis method for planetary gear based on image feature extraction and bag-of-words (BoW) model is proposed, which does not need complex noises reduction and redundant features elimination. In order to extract a large number of features quickly, the FAST-Unoriented-SIFT feature extraction algorithm is proposed, which combines the advantages of FAST and SIFT algorithms, and ignores the orientations of features. After converting the original vibration signals to gray-scale images, the features are extracted directly by FAST-Unoriented-SIFT, and finally the BoW model is built on the extracted features to realize fault classification. The overall recognition rate reaches 98.67%. The experimental results certify the effectiveness of the proposed method.