Search

As deep learning models are usually massive and complex, distributed learning is essential for increasing training efficiency. Moreover, in many real-world application scenarios like healthcare, distributed learning can also keep the data local and protect privacy. Recently, the asynchronous decentralized parallel stochastic gradient descent (ADPSGD) algorithm has been proposed and demonstrated to be an efficient and practical strategy where there is no central server, so that each computing node only communicates with its neighbors. Although no raw data will be transmitted across different local nodes, there is still a risk of information leak during the communication process for malicious participants to make attacks. In this paper, we present a differentially private version of asynchronous decentralized parallel SGD framework, or A(DP) $^2$ 2 SGD for short, which maintains communication efficiency of ADPSGD and prevents the inference from malicious participants. Specifically, Rényi differential privacy is used to provide tighter privacy analysis for our composite Gaussian mechanisms while the convergence rate is consistent with the non-private version. Theoretical analysis shows A(DP) $^2$ 2 SGD also converges at the optimal $\mathcal {O}(1/\sqrt{T})$ O (1 / T) rate as SGD. Empirically, A(DP) $^2$ 2 SGD achieves comparable model accuracy as the differentially private version of Synchronous SGD (SSGD) but runs much faster than SSGD in heterogeneous computing environments. [ABSTRACT FROM AUTHOR]

Short text clustering has far-reaching effects on semantic analysis, showing its importance for multiple applications such as corpus summarization and information retrieval. However, it inevitably encounters the severe sparsity of short text representations, making the previous clustering approaches still far from satisfactory. In this paper, we present a novel attentive representation learning model for shot text clustering, wherein cluster-level attention is proposed to capture the correlations between text representations and cluster representations. Relying on this, the representation learning and clustering for short texts are seamlessly integrated into a unified model. To further ensure robust model training for short texts, we apply adversarial training to the unsupervised clustering setting, by injecting perturbations into the cluster representations. The model parameters and perturbations are optimized alternately through a minimax game. Extensive experiments on four real-world short text datasets demonstrate the superiority of the proposed model over several strong competitors, verifying that robust adversarial training yields substantial performance gains. [ABSTRACT FROM AUTHOR]

Unmanned aerial vehicles (UAVs) have recently attracted tremendous attentions in both industries and academic communities. Thanks to the high mobility and flexibility, UAVs can be deployed in many scenarios to provide various types of services. In these scenarios, the position of the UAVs must be timely and accurately acquired to avoid UAV collisions and realize millimeter-wave beamforming. Particle swarm optimization (PSO) is a potential approach to fulfill localization under GPS-denied environment. However, it has the drawbacks of high complexity and relative large localization error. In this article, we consider the UAV localization problem based on improved PSO, which aims at reducing complexity and localization error. We firstly analyze the performance metrics and performance bounds of conventional PSO in the considered UAV localization scenario. Then, the particle initialization process is reconsidered, where a particle and search space reduction method is introduced as the hierarchical PSO (HPSO). Next, the particle updating schemes are redesigned based on the particle number, where the reference best particle is introduced to deal with the limitations in conventional PSO, this is called reference PSO (RPSO). Lastly, the proposed HPSO and RPSO are validated in simulation results. It is shown that the proposed PSO method has both reduced complexity and localization error compared with conventional PSO and other reference methods. [ABSTRACT FROM AUTHOR]

Data-domain least-squares reverse time migration (DDLSRTM) has been proved to be a more effective imaging tool for complex structures, relative to the standard reverse time migration (RTM) approach. One of the difficulties in DDLSRTM is that the enormous computational costs may impede its application in large-scale 3-D data. To mitigate this problem, with the help of point spread functions (PSFs) and spatial interpolation, we have developed a novel 3-D image-domain least-squares reverse time migration (IDLSRTM) approach, which requires once migration and modeling calculations. However, because of the incomplete acquisition geometry of seismic recordings, IDLSRTM is a highly ill-posed inverse problem. The inverted image from the conventional IDLSRTM approach may suffer from the migration artifacts caused by the coarse source and receiver sampling and spatial discontinuity and instability caused by the truncated PSFs. To solve the ill-posedness and improve image quality, the L1 norm constraint and total variation (TV) regularization are introduced into the objective function of IDLSRTM. The alternating direction method of multipliers (ADMMs) algorithm is developed to solve this optimization problem. Through some 3-D synthetic and field data, it can determine that the proposed IDLSRTM approach computationally efficiently produces a high-fidelity reflection image with good spatial continuity and fewer migration artifacts. It has shown this approach to be a cost-effective and practical inversion-based imaging tool for 3-D field datasets. [ABSTRACT FROM AUTHOR]

This article proposeda SPICE-compatible model to fast simulate the buzz noise problem in a camera device. Based on the reciprocity theorem, the proposed SPICE-compatible model consists of cascaded scattering (S) parameters extracted from the field coupling. It can provide an accurate estimation of the buzz noise transfer function (TFtotal) between the radio-frequency antenna and the audio system within an average error of 1 dB. Besides, the proposed model allows for coupling decomposition by separating the audio system into different regions, fast buzz noise mitigation by adding filtering circuits, and fast antenna evaluation/selection by characterizing the TFtotal with different antenna structures. The advantage of the proposed model is that, instead of multiple 3-D simulations during the predesign-stage troubleshooting, it is a fast SPICE-compatible method with SPICE simulation time of 4.5 min and one-time 3-D full-wave simulation on the camera system studied. It provides quick estimation on the potential buzz noise problems, coupling decomposition,buzz noise mitigation methods, and selection of antenna structures. [ABSTRACT FROM AUTHOR]

User-facing services are now evolving towards the microservice architecture where a service is built by connecting multiple microservice stages. Since the entire service is heavy, the microservice architecture shows the opportunity to only offload some microservice stages to the edge devices that are close to the end users. However, emerging techniques often result in the violation of Quality-of-Service (QoS) of microservice-based services in cloud-edge continuum, as they do not consider the communication overhead or the resource contention between microservices and external co-located tasks. We propose Nautilus, a runtime system that effectively deploys microservice-based user-facing services in cloud-edge continuum. Nautilus ensures the QoS of microservice-based user-facing services while minimizing the required computational resources, which is comprised of a communication-aware microservice mapper, a contention-aware resource manager and an IO-sensitive and load-aware microservice migration scheduler. The mapper divides the microservice graph into multiple partitions based on the communication overhead and maps the partitions to appropriate nodes. On each node, the resource manager determines the optimal resource allocation for its microservices based on reinforcement learning that may capture the complex contention behaviors. Once the microservices are suffered from external IO pressure, the IO-sensitive microservice scheduler migrates the critical one to idle nodes. Furthermore, when the load of microservices changes dynamically, the load-aware microservice scheduler migrates microservices from busy nodes to idle ones to ensure the QoS goal of the entire service. Our experimental results show that Nautilus can guarantee the required QoS target under external shared resources contention while the state-of-the-art suffers from QoS violations. Meanwhile, Nautilus reduces the computational resource usage by 23.9% and the network bandwidth usage by 53.4%, while achieving the required 99%-ile latency. [ABSTRACT FROM AUTHOR]

One of the major challenges facing video object segmentation (VOS) is the gap between the training and test datasets due to unseen category in test set, as well as object appearance change over time in the video sequence. To overcome such challenges, an adaptive online framework for VOS is developed with bi-decoders mutual learning. We learn object representation per pixel with bi-level attention features in addition to CNN features, and then feed them into mutual learning bi-decoders whose outputs are further fused to obtain the final segmentation result. We design an adaptive online learning mechanism via a deviation correcting trigger such that bi-decoders online mutual learning will be activated when the previous frame is segmented well meanwhile the current frame is segmented relatively worse. Knowledge distillation from the well segmented previous frames, along with mutual learning between bi-decoders, improves generalization ability and robustness of VOS model. Thus, the proposed model adapts to the challenging scenarios including unseen categories, object deformation, and appearance variation during inference. We extensively evaluate our model on widely-used VOS benchmarks including DAVIS-2016, DAVIS-2017, YouTubeVOS-2018, YouTubeVOS-2019, and UVO. Experimental results demonstrate the superiority of the proposed model over state-of-the-art methods. [ABSTRACT FROM AUTHOR]

The fact that the millimeter-wave (mmWave) multiple-input multiple-output (MIMO) channel has sparse support in the spatial domain has motivated recent compressed sensing (CS)-based mmWave channel estimation methods, where the angles of arrivals (AoAs) and angles of departures (AoDs) are quantized using angle dictionary matrices. However, the existing CS-based methods usually obtain the estimation result through one-stage channel sounding that have two limitations: (i) the requirement of large-dimensional dictionary and (ii) unresolvable quantization error. These two drawbacks are irreconcilable; improvement of the one implies deterioration of the other. To address these challenges, we propose, in this paper, a two-stage method to estimate the AoAs and AoDs of mmWave channels. In the proposed method, the channel estimation task is divided into two stages, Stage I and Stage II. Specifically, in Stage I, the AoAs are estimated by solving a multiple measurement vectors (MMV) problem. In Stage II, based on the estimated AoAs, the receive sounders are designed to estimate AoDs. The dimension of the angle dictionary in each stage can be reduced, which in turn reduces the computational complexity substantially. We then analyze the successful recovery probability (SRP) of the proposed method, revealing the superiority of the proposed framework over the existing one-stage CS-based methods. We further enhance the reconstruction performance by performing resource allocation between the two stages. We also overcome the unresolvable quantization error issue present in the prior techniques by applying the atomic norm minimization method to each stage of the proposed two-stage approach. The simulation results illustrate the substantially improved performance with low complexity of the proposed two-stage method. [ABSTRACT FROM AUTHOR]

This article proposes a novel surrogate-assisted multistate tuning-driven electromagnetic (EM) optimization technique to address the challenges of microwave tunable filter design with multiple tuning states. The desired multiple tuning states are satisfied simultaneously using the proposed surrogate-assisted technique. The proposed surrogate model is composed of several subsurrogate models. Each subsurrogate model is developed to perform the optimization for each tuning state. The subsurrogate models share the same values of nontunable parameters and possess different values of tunable parameters. The overall surrogate model is developed to find a single set of optimal solutions for nontunable parameters and multiple sets of optimal solutions for tuning parameters simultaneously. Parallel computation scheme is exploited to generate the training samples for establishing the proposed surrogate model. Furthermore, a new trust-region updating formulation specifically for multistate tuning is proposed to improve the convergence of the proposed optimization algorithm. Using the proposed optimization technique, different tuning states are considered together and optimized simultaneously. The values of nontunable design parameters are constrained by all tuning states and consequently there is a higher chance that more suitable solutions can be found to satisfy all the desired tuning states simultaneously. The proposed technique for the tunable filter design with multiple tuning states has a better capability of avoiding local minima and can reach the optimal solution more effectively in comparison with the existing optimization method. Two microwave examples are used to validate the proposed technique. [ABSTRACT FROM AUTHOR]

Datacenters use GPUs to provide the significant computing throughput required by emerging user-facing services. The diurnal user access pattern of user-facing services provides a strong incentive to co-located applications for better GPU utilization, and prior work has focused on enabling co-location on multicore processors and traditional non-preemptive GPUs. However, current GPUs are evolving towards spatial multitasking and introduce a new set of challenges to eliminate QoS violations. To address this open problem, we explore the underlying causes of QoS violation on spatial multitasking GPUs. In response to these causes, we propose C-Laius, a runtime system that carefully allocates the computation resource to co-located applications for maximizing the throughput of batch applications while guaranteeing the required QoS of user-facing services. C-Laius not only allows co-locating one user-facing application with multiple batch applications, but also supports the co-location of multiple user-facing applications with batch applications. In the case of a single co-located user-facing application, our evaluation on an Nvidia RTX 2080Ti GPU shows that C-Laius improves the utilization of spatial multitasking GPUs by 20.8 percent, while achieving the 99%-ile latency target for user-facing services. As to the case of multiple co-located user-facing applications, C-Laius ensures no violation of QoS while improving the accelerator utilization by 35.9 percent on average. [ABSTRACT FROM AUTHOR]

Distributed phased Multiple-Input Multiple-Output (phased-MIMO) radar system consisting of separated subarrays possesses a high angle resolution but with sharing high sidelobes. This paper therefore focuses on the design of a virtual low-sidelobe beampattern in distributed phased-MIMO radar. An iterative framework with respect to subarray layout and receive weighting coefficients is proposed to minimize beampattern Peak Sidelobe Level (PSL) accounting for practical position constraints as well as the mainlobe level restriction. In each iteration, an efficient cyclic optimization algorithm based on the Coordinate Descent (CD) framework is developed to seek for the subarray layout through splitting high dimensional layout optimization problem into multiple one-dimensional optimization problems, and the SemiDefinite Relaxation (SDR) technique and rank one approximation are explored to design weighting coefficients. Numerical simulations are provided to assess the performance of the proposed algorithms in term of the achieved beampattern under different constraints and parameter settings. [ABSTRACT FROM AUTHOR]

Using deep neural networks (DNNs) to explore spatial patterns and temporal dynamics of human brain activities has been an important yet challenging problem because the artificial neural networks are hard to be designed manually. There have been several promising deep learning methods, e.g., deep belief network (DBN), convolutional neural network (CNN), and deep sparse recurrent auto-encoder (DSRAE), that can decompose neuroscientific and meaningful spatiotemporal patterns from 4D functional Magnetic Resonance Imaging (fMRI) data. However, those previous studies still depend on hand-crafted neural network architectures and hyperparameters, which are not optimal in various senses. In this paper, we employ the evolutionary algorithms (EA) to optimize the deep neural architecture of DSRAE by minimizing the expected loss of initialized models, named eNAS-DSRAE (evolutionary Neural Architecture Search on Deep Sparse Recurrent Auto-Encoder). Also, validation experiments are designed and performed on the publicly available human connectome project (HCP) 900 datasets, and the results achieved by the optimized eNAS-DSRAE suggested that our framework can successfully identify the spatiotemporal features and perform better than the hand-crafted neural network models. To our best knowledge, the proposed eNAS-DSRAE is not only among the earliest NAS models that can extract connectome-scale meaningful spatiotemporal brain networks from 4D fMRI data, but also is an effective framework to optimize the RNN-based models. [ABSTRACT FROM AUTHOR]

Automatic fracture identification and segmentation in well logging images is increasingly critical but arduous because of extensive exploration for oil and gas. Domain adaptation for semantic segmentation is an appealing alternative to the lack of semantic fracture annotations. Nevertheless, previous domain adaptation strategies merely focus on a single level (e.g., input, intermediate features, or output), which could result in a large generalization error on new well logging images. In this article, a two-stage network architecture is proposed. In stage 1, a pattern transfer network (PTN) model is utilized to carry out a transformation from one domain to the other domain, making images visually similar between two datasets. Sequentially, in stage 2, an adversarial learning model with a generator and discriminator is introduced to make two domain images generated via PTN producing similar semantic segmentation. Furthermore, an attention module was embedded in the generator to filter irrelevant noise in semantic segmentation results. Real logging images were collected as the tested datasets by an ultrasonic imaging logging instrument. Compared to earlier conventional algorithms and domain adaptation methods, the proposed method is shown to obtain better visual quality and accurate segmentation outcomes. [ABSTRACT FROM AUTHOR]

Employing grounded theory methodology, this article investigates collaborative innovation based on interview and secondary data collected from a technology alliance. The findings reveal that alliance-specific capabilities and knowledge-based capabilities are the two critical drivers. Alliance-specific capabilities include such components as alliance management, mutual learning, and alliance innovation; while knowledge-based capabilities refer to knowledge-centered activities of exploration, exchange, integration, and exploitation. Further, the two types of capabilities interact with each other in creating a technological knowledge base through mobilizing knowledge and in, ultimately, producing technological innovation. Taken together, this article extends the existing literature of technology alliance by providing a knowledge-based, capability-building view in unpacking collaborative innovation. It also provides implications for managers who aim to enhance technological innovation via alliance cooperation. [ABSTRACT FROM AUTHOR]

In this paper, we tackle the problem of pose-guided person image generation with unpaired data, which is a challenging problem due to non-rigid spatial deformation. Instead of learning a fixed mapping directly between human bodies as previous methods, we propose a new pathway to decompose a single fixed mapping into two subtasks, namely, semantic parsing transformation and appearance generation. First, to simplify the learning for non-rigid deformation, a semantic generative network is developed to transform semantic parsing maps between different poses. Second, guided by semantic parsing maps, we render the foreground and background image, respectively. A foreground generative network learns to synthesize semantic-aware textures, and another background generative network learns to predict missing background regions caused by pose changes. Third, we enable pseudo-label training with unpaired data, and demonstrate that end-to-end training of the overall network further refines the semantic map prediction and final results accordingly. Moreover, our method is generalizable to other person image generation tasks defined on semantic maps, e.g., clothing texture transfer, controlled image manipulation, and virtual try-on. Experimental results on DeepFashion and Market-1501 datasets demonstrate the superiority of our method, especially in keeping better body shapes and clothing attributes, as well as rendering structure-coherent backgrounds. [ABSTRACT FROM AUTHOR]

Instead of streaming the entire omnidirectional video that induces a waste of bandwidth, the tile-based viewport adaptive streaming is preferred in practice. Tiles within users’ viewport are transmitted in high quality, while tiles elsewhere are delivered in lower quality to reduce the network bandwidth consumption. In this setup, when users change their viewport, they may potentially be disturbed by the exposure of low-quality tiles and the subsequent quality refinement in a period of delay. We defined this phenomenon in the viewport as a Dynamic Quality Boundary (DQB) artifact, as there exists a clear boundary between low-quality tiles and high-quality tiles which appears, moves, and disappears along with the change of users’ viewport. The DQB artifact is a specific quality degradation in the viewport-adaptive omnidirectional video streaming whose impact on users’ perceptual quality has not been adequately studied. In this paper, we focused on investigating and modeling its impact on the quality perception of users when streaming omnidirectional videos. More specifically, systematic subjective quality evaluation experiments towards DQB artifact were designed and conducted. Accordingly, a perceptual quality metric was proposed considering the quality of tiles, the proportion of low-quality tiles, and the refinement duration. Experimental results, measured in various aspects, showed that the proposed metric can accurately predict users’ perceptual quality of the tile-based viewport-adaptive omnidirectional video streaming service. [ABSTRACT FROM AUTHOR]

Modern processors often need to switch among different power states based on usage scenarios, energy availability, and thermal conditions. Dynamic voltage and frequency scaling (DVFS) is a commonly used power management strategy to trade off performance and energy. As transistor scaling is reaching its limit, the viable supply voltage range where DVFS can operate is shrinking, which limits its effectiveness. To extend the performance-energy trade-off capabilities in modern processors, this article proposes dynamic core scaling (DCS) that does not rely on voltage scaling. DCS dynamically adjusts the active superscalar datapath resources so that programs run at a given percentage of their maximum speed while minimizing energy consumption at the same time. Since DCS does not need voltage scaling, it can be combined with DVFS to achieve greater energy savings. To effectively manage performance-energy trade-offs using a combination of DCS and DVFS, this article proposes an oracle controller that demonstrates the optimal control strategy, and two practical controllers that are applicable in real implementations. Evaluations using an 8-way superscalar processor implemented in a 45-nm circuit show that DCS is more effective in performance-energy trade-offs than DVFS at the high performance end for a number of SPEC CPU2000 benchmarks. When used together with DVFS, DCS saves an additional 20 percent of a full-size core’s energy on average. At the minimum operating voltage, DVFS hits its limit, while DCS is still able to achieve an average of 46 percent further energy reduction. [ABSTRACT FROM AUTHOR]

Seismic full-waveform inversion (FWI) aims to build high-resolution images of the physical properties of the subsurface. However, the ill-posedness and nonlinear problems pose a great challenge to the high-resolution reconstruction. Although the nonlinear problem can be mitigated by matching a subset of observation data, the resulting images are generally low-resolution background structures. Regularization-based techniques can mitigate the ill-posedness of FWI, but the iterative method suffers from the cycle-skipping and computational burden problems. To overcome these problems, we develop an adjoint-driven deep-learning FWI (AD-DLFWI) approach which utilizes the fully convolutional network (FCN) to invert subsurface velocity from reflection seismic data. Specifically, AD-DLFWI is implemented in a two-step iterative scheme, in which an optimal update result at each step is learned via a FCN-based learned updating operator. The proposed approach uses the seismic image of applying the adjoint operator of the scattering wave equation, which is equivalent to the gradient of classical FWI, as the data engine of FCN. Inspired by the wave-equation migration velocity analysis approach, we propose to unfold the gradient of FWI into the common-source domain to keep the information about the measure of velocity error. To ensure the interpretability of each network’s role, we design a two-step training scheme to successively reconstruct the low and high wavenumber components of subsurface velocity. Using synthetic experiments with reflection-dominant seismic data, we have confirmed that the proposed FWI approach not only can provide a reliable velocity estimation but also is not sensitive to the cycle-skipping problem. [ABSTRACT FROM AUTHOR]

Process fluctuations can significantly affect the yield of millimeter-wave circuits based on submicrometer semiconductor processes. A yield-improvement design that incorporates passive circuits is well established. However, how to consider the process fluctuation in active circuit designs (e.g., power amplifiers) and improve yield is still a challenge. In this article, a yield-improvement load—pull design method, which can be used to find optimum load impedance accounting for high yield while maintaining good output performance, is presented. The yield-improvement load—pull theory is described using an analytical model of transistor large-signal performance considering the process fluctuation. A method of the yield-improvement load—pull for the practical microwave monolithic integrated circuit (MMIC) design is introduced in detail. For demonstration purposes, a high-performance Ka-band power amplifier MMIC is designed. The results show that with the proposed yield-improvement load—pull method, an obvious improvement of yield can be observed with only minimal sacrifice of the output power (Pout) and the power added efficiency (ηPAE), respectively. As a result, the method of this article can be used to improve the yield of millimeter-wave/terahertz integrated circuits (ICs). [ABSTRACT FROM AUTHOR]

Superpixels are widely used in computer vision applications. Most of the existing superpixel methods use established criteria to indiscriminately process all pixels, resulting in superpixel boundary adherence and regularity being unnecessarily inter-inhibitive. This study builds upon a previous work by proposing a new segmentation strategy that classifies image content into meaningful areas containing object boundaries and meaningless parts that include color-homogeneous and texture-rich regions. Based on this classification, we design two distinct criteria to process the pixels in different environments to achieve highly accurate superpixels in content-meaningful areas and keep the regularity of the superpixels in content-meaningless regions. Additionally, we add a group of weights when adopting the color feature, successfully reducing the undersegmentation error. The superior accuracy and the moderate compactness achieved by the proposed method in comparative experiments with several state-of-the-art methods indicate that the content-adaptive criteria efficiently reduce the compromise between boundary adherence and compactness. [ABSTRACT FROM AUTHOR]

Errors induced by pillar taper and eccentric load in micropillar compression experiments have seldom been investigated, although they significantly affect the measurement accuracy, especially in the testing of specimens in a micrometer and nanometer size. A novel calibration method related to stress–strain responses in micropillar compression testing was proposed in this study to eliminate this kind of errors. Axial load and eccentric load were applied to high-entropy alloy (HEA) pillars with taper angles of 1°, 2°, and 3°, respectively, and the measured stress–strain curves were corrected by the proposed calibration method. The results revealed that the difference of the corrected curves with different tapers or eccentricities was smaller than the measured curves. The errors induced by the taper and eccentric load increased with the taper angle and eccentricity. In addition, the difference between measured and corrected flow stresses increased with the taper angle, which was essentially consistent with the results from the calibration method. [ABSTRACT FROM AUTHOR]

Reconfigurable Intelligent surface (RIS), which can induce phase shift of the received signals and then forward them by its reflecting elements, possesses high energy efficiency (EE). This letter studies the EE for the RIS-assisted multicast communication. A multiple-antenna base station (BS) delivers common messages to a group of mobile users (MUs), The RIS is placed to maintain the transmission. First, this letter analyzes the equivalent multicast channel and total power consumption to formulate an EE maximization problem. Then, this letter proposes the functional algorithms by fractional programming and alternating optimization to obtain a locally optimal solution for this problem. Lastly, we show asymptotic behaviors of the EE when the numbers of reflecting elements, antennas, and MUs go to infinity. [ABSTRACT FROM AUTHOR]

In the past years, deep learning-based machinery fault diagnosis methods have been successfully developed, and the basic diagnostic problems have been well addressed where the training and testing data are collected under the same operating conditions. When the training and testing data are from different distributions, domain adaptation approaches have been introduced. However, the existing methods generally assume the availability of the target-domain data in all the health conditions during training, which is not in accordance with the real industrial scenarios. This article proposes a deep learning-based fault diagnosis method to address the partial domain adaptation problems, where the unsupervised target-domain training data do not cover the full machine health state label space. The conditional data alignment and unsupervised prediction consistency schemes are proposed to achieve partial domain adaptation. The experimental results on two rotating machinery datasets suggest the proposed method offers a promising tool for this practical industrial problem. [ABSTRACT FROM AUTHOR]

Space mapping is a recognized surrogate-based optimization method to accelerate the electromagnetic (EM) design. In this article, for the first time, space mapping is elevated from solving the problem of EM optimization to solving the problem of multiphysics optimization for high power microwave filters. Multiphysics analysis, which involves the EM domain with other physics domains, is increasingly important for high-performance microwave components to obtain an accurate system design. To speed up the multiphysics design, a space-mapping-based surrogate model including a coarse model and two mapping functions is proposed in this article. We propose to use EM single physics responses as the coarse model to provide good approximations to fine model multiphysics responses. To avoid repetitive EM simulations during the surrogate model training and optimization process, the coarse model is developed using an artificial neural network (ANN). Frequency mapping and explicit input mapping are further performed to develop the proposed surrogate model. Multiple EM and multiphysics training samples are evaluated in parallel to develop the surrogate model. A trust-region algorithm, tailored to the space-mapping-based multiphysics optimization technique, is proposed to improve the convergence. By exploiting the knowledge of the coarse model established by relatively inexpensive EM data, the proposed technique can provide a larger and more efficient optimization update in each optimization iteration, consequently obtaining optimal solutions faster than the existing multiphysics optimization without space mapping. Two examples of multiphysics optimization of high-power microwave filters are used to validate the proposed technique. [ABSTRACT FROM AUTHOR]

Early warning of a potential pandemic with respiratory symptoms is crucial for global health management. It enables timely intervention to reduce the likelihood of uncontrollable massive virus spread. In this research, we propose to leverage the ubiquitous wearable devices to develop a wearable crowdsource system to monitor respiratory symptoms such as cough and fever. Wearable devices nowadays can directly and non-intrusively measure people's vital signs in real time with a variety of sensors embedded. We collect the data from wearable devices and develop machine learning algorithms to analyze the data for respiratory symptom monitoring and early warning. In particular, we focus on cough detection through multi-source data fusion (e.g., accelerometer amplitude and microphone audio). Preliminary results show that our algorithms result in higher detection accuracy and less false positives with the lowest use of computing resources. This research potentially transforms the way pandemic early warning is implemented and the way we respond to public health crises in the years to come. [ABSTRACT FROM AUTHOR]

Highly sensitive and cost-effective flexible piezoresistive sensors are pursued as essential components of electronic skin (e-skin) for a variety of applications such as soft robotics and body prosthesis. A common strategy for achieving high sensitivity of these sensors is to fabricate their electrode material with complex 3-D microstructures, but manufacturing cost-effective micro-structured sensors with high reliability and reproducibility is a challenge. Herein, a robust approach for producing high-reproductive microstructured polydimethylsiloxane (PDMS) films composing the sensors is proposed, based on silicon molds with a honeycomb-like architecture fabricated rapidly and cost-effectively at a standoff distance by femtosecond laser pulses in the self-channeling regime. Single-walled carbon nanotubes (SWNTs) are embedded in the micro-structured PDMS films composing sandwich-structural flexible piezoresistive sensors. The sensing devices exhibit good performance with high sensitivity, fast response time, and excellent stability and are used with success for the detection of hand motion pressure and blood pressure at the wrist, thus showing a great potential as components of e-skin for detection of various human motions. [ABSTRACT FROM AUTHOR]

In this paper, the problem of describing visual contents of a video sequence with natural language is addressed. Unlike previous video captioning work mainly exploiting the cues of video contents to make a language description, we propose a reconstruction network (RecNet) in a novel encoder-decoder-reconstructor architecture, which leverages both forward (video to sentence) and backward (sentence to video) flows for video captioning. Specifically, the encoder-decoder component makes use of the forward flow to produce a sentence based on the encoded video semantic features. Two types of reconstructors are subsequently proposed to employ the backward flow and reproduce the video features from local and global perspectives, respectively, capitalizing on the hidden state sequence generated by the decoder. Moreover, in order to make a comprehensive reconstruction of the video features, we propose to fuse the two types of reconstructors together. The generation loss yielded by the encoder-decoder component and the reconstruction loss introduced by the reconstructor are jointly cast into training the proposed RecNet in an end-to-end fashion. Furthermore, the RecNet is fine-tuned by CIDEr optimization via reinforcement learning, which significantly boosts the captioning performance. Experimental results on benchmark datasets demonstrate that the proposed reconstructor can boost the performance of video captioning consistently. [ABSTRACT FROM AUTHOR]

This paper presents a secure chaotic communication scheme using observer-based synchronization of chaotic systems. A class of chaotic systems satisfying incremental quadratic constraints are considered as the transmitter. The original message is encrypted in the noise-like output of the system. A circle criterion-based observer is designed for the system as the receiver by using the linear matrix inequalities technique. The message is recovered at the receiver end from the estimation error of the observer. To design the observer, a general algorithm to calculate incremental multiplier matrices for nonlinearities in chaotic systems is presented. Moreover, the proposed observer-based chaotic secure communication scheme is feasible for many common chaotic systems with ellipsoidal bounds. The simulation results show the good performance of the proposed communication scheme through an example of image transmission. [ABSTRACT FROM AUTHOR]

This article proposes a novel decomposition technique to address the challenges of electromagnetic (EM) parametric modeling, where the values of geometrical parameters change in a large range. In this method, a systematic and automated algorithm based on second-order derivative information is proposed to decompose the overall geometrical range into a set of subranges. Using the proposed technique, a smooth region is decomposed into a few large subregions, while a highly nonlinear region is decomposed into many small subregions. The proposed technique provides an efficient mathematical methodology to perform the decomposition in a systematic and automated process. An artificial neural network (ANN) model with a simple structure, hereby referred to as a submodel, is developed with geometrical parameters as variables in each subregion. When the values of geometrical parameters change from the region of one submodel to another submodel, the discontinuity of the EM responses is observed at the boundary between the adjacent submodels. There are many submodel boundaries in the overall model resulting in the complex multidimensional discontinuity problem. A submodel modification process is proposed to solve this multidimensional discontinuity problem to obtain a continuous model over the entire region. Parallel data generation, parallel submodel training, and parallel submodel modification are proposed to speed up the modeling development process. Compared with standard modeling methods using a single model to cover the entire wide geometrical range, the proposed method can obtain better model accuracy with short model-development time. Two microwave filter examples are used to illustrate the validity of the proposed technique. [ABSTRACT FROM AUTHOR]

Multi-panel MIMO is a practical antenna array structure for massive MIMO, but the optimal hybrid beamforming design has a complexity exponential to the number of antenna panels. In this letter, we propose a truncated tree based hybrid beamforming design with linear complexity for multi-panel MIMO transmission. We convert the hybrid beamforming design problem into the shortest path finding problem that is visualized by tree data structure. To reduce complexity, the tree is truncated by deleting the intermediate paths that are less likely to be the shortest path. By this means, the truncated tree based design reduces exponential time complexity to linear time complexity. Numerical results show that the proposed truncated tree based design far outperforms the heuristic max power gain beamforming design and achieves near optimal performance with a proper truncation ratio. [ABSTRACT FROM AUTHOR]

Radio frequency (RF) signal based wireless power transfer (WPT) is incorporated into fog radio access network (F-RAN) in order to simultaneously satisfy the content downloading demand of the content users (CUs) and the charging requirement of the energy users (EUs). Numerous enhanced remote-radio-heads (eRRHs) of F-RAN are distributed in a wide area, which are connected to the central base-band-unit (BBU) pool by fronthaul links. In order to reduce the adverse effect of the limited fronthaul capacity on the wireless content transfer (WCT), eRRHs are equipped with the content caching capability. A greedy algorithm based near-optimal caching design is then proposed for reducing the content responsive latency by considering the geographic distribution of CUs. Furthermore, given this caching design, a joint optimisation problem of eRRH classification (some for WCT and some for dedicated WPT) and distributed transmit beamforming is solved by a low-complexity heuristic algorithm. This algorithm aims for maximising the total energy harvested by the EUs, while satisfying the CUs’ minimum WCT rate and the limited fronthaul capacity. Intensive simulation results in the scenarios of both uniform and non-uniform users’ distribution validate our caching, eRRH classification and distributed transmit beamforming design in an F-RAN. [ABSTRACT FROM AUTHOR]

Energy harvesting promises to power billions of Internet-of-Things devices without being restricted by battery life. The energy output of harvesters is typically tiny and highly unstable, so the computing system must store program states into nonvolatile memory frequently to preserve the execution progress in the presence of frequent power failures. Task-based intermittent computing is a promising paradigm to provide such capability, where each task executes atomically and only states across task boundaries need to be saved. This article presents LATICS, a low-overhead adaptive task-based intermittent computing system, which dynamically decides the granularity of atomic execution to avoid unnecessarily frequent state saving when energy supply is sufficient. The novel feature of LATICS is to drastically reduce the amount of states to be saved at task boundaries compared with existing solutions. Notably, we disclose that skipping state saving at some task boundary may cause the system to store more states at other places, and thus leads to higher overall overhead. Therefore, LATICS enforces mandatory state saving at certain task boundaries regardless of the current energy condition to reduce state saving overhead. We implement LATICS on a real energy-harvesting platform based on MSP430 and experimentally compare against the state-of-the-art under different settings. The experimental results show that LATICS significantly reduces state saving overhead and improves execution efficiency compared to existing solutions. [ABSTRACT FROM AUTHOR]

Visual object tracking is a fundamental and time-critical vision task. However, most trackers such as SiamFC and CFNet missed the object movement and simply defined the searching region centered at the location of the target in the previous frame. So they tend to fail in the cases with severe occlusion or a large displacement of the target. In this paper, we consider the object tracking as a dual-task problem of guessing and matching. A guess module is to estimate the motion trend of the target by reinforcement learning based on the observations on appearance changes and motion history. Rather than using the previous location of the target, we may have a more accurate center to locate the searching region. Benefited from such improved searching region, the match module becomes less prone to the object drift problem, and can easily identify the target from the potential distractors in the background. Extensive experimental results on benchmark datasets such as RGBT, OTB-2013, OTB-50 and OTB-100, show that the proposed method achieves leading performance compared to state-of-the-art trackers. Moreover, the proposed tracker could maintain real-time speed, giving itself the potential in practical applications. [ABSTRACT FROM AUTHOR]

Timing analysis of real-time systems must consider cache-related pre-emption delay (CRPD) costs when pre-emptive scheduling is used. While most previous work on CRPD analysis only considers instruction caches, the CRPD incurred on data caches is actually more significant. The state-of-the-art CRPD analysis methods are based on useful cache block (UCB) calculation. Unfortunately, as shown in this article, directly extending the existing UCB calculation techniques from instruction caches to data caches will lead to both unsoundness and significant imprecision. To solve these problems, we develop a new UCB calculation technique for data caches, which redefines the analysis unit (to address the unsoundness in the existing method) and precisely captures the dynamic cache access behavior by taking the temporal scopes of memory blocks into consideration. The experimental results show that our new technique yields substantially tighter CRPD estimations comparing with the state-of-the-art. [ABSTRACT FROM AUTHOR]

Image reconstruction for ultrasonic transmission tomography (UTT) is a complex problem. Due to the limited diffusion angle of ultrasonic transducers, few useful data are received from the opposite transducers so the image reconstruction problem is ill-posed. The traditional logic backprojection (LBP) algorithm cannot sufficiently deal with the sparse characteristic of measurement data. In this article, transducers with a large diffusion angle are designed to collect more information. A sparse preimaged orthogonal matching pursuit (SP-OMP) algorithm is presented for UTT image sparse reconstruction. The performance of the proposed algorithm is experimentally evaluated by a UTT system versus classical algorithms such as LBP, algebra reconstruction technique (ART), simultaneous iteration reconstruction (Sirt), Tikhonov, and L1-regularization. The quality parameters such as mean square error (MSE), position error (PE), size error (SE), relative error (RE), and correlation coefficient (CC) are surveyed which indicated that the SP-OMP algorithm could provide higher quality images than those provided by the conventional image reconstruction algorithms. [ABSTRACT FROM AUTHOR]

In the recent years, data-driven machinery fault diagnostic methods have been successfully developed, and the tasks where the training and testing data are from the same distribution have been well addressed. However, due to sensor malfunctions, the training and testing data can be collected at different places of machines, resulting in the feature space with significant distribution discrepancy. This challenging issue has received less attention in the current literature, and the existing approaches generally fail in such scenarios. This article proposes a domain adaptation method for machinery fault diagnostics based on deep learning. Adversarial training is introduced for marginal domain fusion, and unsupervised parallel data are explored to achieve conditional distribution alignments with respect to different machine health conditions. Experiments on two rotating machinery datasets are carried out for validations. The results suggest the proposed method is promising to address the fault diagnostic tasks with data from different places of machines, further enhancing applicability of data-driven methods in real industries. [ABSTRACT FROM AUTHOR]

Frequency modulated continuous wave (FMCW) based millimeter wave (MMW) radar systems have found widespread applications in advanced driving assistant system recently. MMW radar has clear superiority in many aspects, such as excellent range resolution, adaptability to various environments and low cost. However, its fatal weakness is the poor angular resolution which is very important in automotive radar applications. This shortcoming can be overcome by using a physically large aperture. However, achieving moderate angular resolution requires an array that is too large to be practically integrated into a vehicle. In this paper, a high resolution direction of arrival (DOA) approach is proposed by utilizing the relative motion between radar and targets, and a very large virtual aperture can be created by combining the signal vectors from multiple consecutive coherent processing intervals (CPIs). It should be pointed out that, the proposed approach is applicable for both stationary and moving targets even without motion data of vehicles. Therefore, it is completely different from the synthetic aperture radar (SAR). The effectiveness is verified by both simulation results and the measurement data. It is demonstrated that the angular resolution can be superior to 1° by a linear array consisting of only 4 elements. [ABSTRACT FROM AUTHOR]

Unmanned Aerial Vehicles (UAVs) have the advantages of high mobility and flexibility that can be potentially used as aerial base stations (BSs) to provide millimeter-wave (mmWave) communication with ground users. However, the UAV mmWave beamforming must be carefully designed to provide directional transmission and support reliable connections. In this paper, we consider a UAV mmWave system, where the UAV serves as a BS to cover mobile users on the ground. Firstly, we study the beam training and propose beam training codebook designs for both insufficient and sufficient beam training cases. Then, we propose efficient beam tracking methods to predict the beam direction with user mobility, where both random and inertial motion are considered. Simulation results show that the average downlink capacity increases with the proposed beam training codebook. Further, the mean angular error of our proposed beam tracking is lower than existing method. [ABSTRACT FROM AUTHOR]

In a multimodule system, an increase in the number of radiating optical modules will increase the electromagnetic emissions. This article investigates the scaled tendency of the emissions in a router system loaded with hundreds of optical modules. An artificial router mimicking the real system was built to investigate this tendency. A patch antenna array mimics the radiation of the optical modules. It can be excited in in-phase and random-phase configurations. The measurement data verifies the theoretical analysis and the prediction from a statistical method without performing hundreds of different experiments on a real router. Assuming that all radiators are radiating at the same frequency and have similar radiation pattern with random phases, the average of the maximal directivity of the system will saturate if the number of radiators (N) is larger than 14. Furthermore, the average of the maximal electric field radiated will increase following a 10 log10 N (dB) tendency. [ABSTRACT FROM AUTHOR]

Beamforming is an enabling technique for wireless communications over higher frequency band, e.g., mmWave frequency band. To enhance beam concentration and homogenization within any angle range, we propose a hybrid beamforming design in three steps. First, through analyzing radiation power ratio, we find the bases of beamforming vector to guarantee a satisfactory level of power radiated over the targeted angle range. Second, with the derived bases, we flat the response of the passband by solving a max-min optimization problem using semidefinite relaxation technique. Third, we perform hybrid decomposition to achieve compatibility with phased array antenna with hybrid structure. The proposed beamforming is verified through quantitative measurement of beam concentration and homogenization. Simulation results show that the beamform pattern designed by our method outperform other existing methods in multi-cell scenarios. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel deep neural network based attention model to learn the representative local regions from a video sequence for person re-identification. Specifically, we propose a multi-scale spatial-temporal attention (MSTA) model to measure the regions of each frame in different scales from the perspective of whole video sequence. Compared to traditional temporal attention models, MSTA focuses on exploiting the importance of local regions of each frame to the whole video representation in both spatial and temporal domains. A new training strategy is designed for the proposed model by incorporating the image-to-image mode with the video-to-video mode. Extensive experiments on benchmark datasets demonstrate the superiority of the proposed model over state-of-the-art methods. [ABSTRACT FROM AUTHOR]

The ac power supply system of the magnetic island divertor coil is widely applied to reduce the peak thermal load of the boundary target plate. It was found that this system should have the current frequency in the range of 0–100 Hz and the amplitude of 5 kA at 100 Hz. In this article, a new main circuit is proposed for the variable-frequency power supply system. The proposed circuit consists of a six-pulse, thyristor rectifier and a parallel, H-bridge inverter. The control strategy applies the carrier phase-shifted, sinusoidal pulsewidth modulation (CPS-SPWM) to form an almost sinusoidal current and calculate the LC filter parameters of the bus and inverter. Based on the requirements and load parameters of the power supply, it is found that the voltage and current double-closed loop PI control accelerates the system dynamic response and improves the nonlinear, load adaptability of the system. Moreover, the phase-shifting transformer with high-overload capability is applied in the proposed power supply to reduce the volume of the transformer and in advance charging. On the other hand, the water-cooling mode is adopted and the structure of water-cooled plate is designed. Moreover, optimal device and layout selection is performed to achieve reasonable modularization. Finally, the proposed method is applied to complete the main components of the space layout of the power supply. In the studied cases, the power density of 0.42 W/L is obtained. [ABSTRACT FROM AUTHOR]

Data transmission over the millimeter wave (mmWave) in fifth-generation wireless networks aims to support very high speed wireless communications. A substantial increase in spectrum efficiency for mmWave transmission can be achieved by using advanced hybrid analog-digital precoding, for which accurate channel state information (CSI) is the key. Rather than estimating the entire channel matrix, it is now well-understood that directly estimating subspace information, which contains fewer parameters, does have enough information to design transceivers. However, the large channel use overhead and associated computational complexity in the existing channel subspace estimation techniques are major obstacles to deploy the subspace approach for channel estimation. In this paper, we propose a sequential two-stage subspace estimation method that can resolve the overhead issues and provide accurate subspace information. Utilizing a sequential method enables us to avoid manipulating the entire high-dimensional training signal, which greatly reduces the computational complexity. Specifically, in the first stage, the proposed method samples the columns of channel matrix to estimate its column subspace. Then, based on the obtained column subspace, it optimizes the training signals to estimate the row subspace. For a channel with $N_r$ receive antennas and $N_t$ transmit antennas, our analysis shows that the proposed technique only requires $\mathcal {O}(N_t)$ channel uses, while providing a guarantee of subspace estimation accuracy. By theoretical analysis, it is shown that the similarity between the estimated subspace and the true subspace is linearly related to the signal-to-noise ratio (SNR), i.e., $\mathcal {O}(\text{SNR})$ , at high SNR, while quadratically related to the SNR, i.e., $\mathcal {O}(\text{SNR}^2)$ , at low SNR. Simulation results show that the proposed sequential subspace method can provide improved subspace accuracy, normalized mean squared error, and spectrum efficiency over existing methods. [ABSTRACT FROM AUTHOR]

Multi-panel MIMO is a promising technology in millimeter wave communications. Due to its partially hybrid structure and non-uniform antenna array, existing channel estimation cannot be directly applied to multi-panel MIMO. In this paper, we study channel estimation for multi-panel mmWave MIMO. We firstly model channel estimation of multi-panel MIMO as a block-sparse signal recovery problem. Then, according to maximum likelihood criterion, we propose an orthogonal projection method to firstly detect the support of the sparse channel response vector and then perform least square estimation, based on which we also propose a low-complexity greedy support detection in order to reduce computational complexity. Finally, through analyzing pairwise error probability of support detection, we verify the performance gain of joint multi-panel channel estimation over single panel channel estimation, and we further find that joint multi-panel channel estimation with independently generated random combining matrices outperforms that with identically generated combining matrices. Numerical results verify the superiority of our proposed joint multi-panel orthogonal projection based channel estimation over existing schemes. [ABSTRACT FROM AUTHOR]

A major assumption in data mining and machine learning is that the training set and test set come from the same domain. They share the same feature space and have the same distribution. However, in many real-world applications, the training set and test set usually come from different domains. Thus, there might be negative similarities between different domains so that the negative transfer problem caused by negative similarity may happen. In this paper, we propose a novel method named active transfer learning (ATL) to solve the above problem. Specifically, the orthogonal projection matrix and the weight coefficient vector are introduced to extend maximum mean discrepancy (MMD) so that it can minimize MMD and simultaneously eliminate the negative transfer. To find the informative and discriminative subsets from the source domain, we then propose an information diversity term by using the local geometric structure information of the source samples. Besides, by using the label information of source samples, our method can guarantee the selected subsets as discriminative as possible. Finally, to efficiently implement the proposed method, an alternating optimization approach, which is based on the alternating direction method of multipliers (ADMM), is designed to solve the optimization problem. To demonstrate the effectiveness of the proposed ATL model, experiments are conducted on five real-world data sets. The experimental results show the superiority of our method over the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Typical networking equipment contains tens of nearly identical line cards and hundreds of the optical modules. The more modules and line cards are added, the larger the radiation will be. This paper presents an idea and approach to estimate the emission of a fully populated system based on data that are obtained using an only partially populated system. [ABSTRACT FROM AUTHOR]

Energy management is a critical challenge in multicore processors due to continuous technology scaling. Previous methods have mostly focused on the energy minimization of the processor cores. However, energy overhead of the off-chip voltage regulator (VR) has recently shown to be a nontrivial part of the total energy consumption and has been previously overlooked. In this paper, we propose an overall energy optimization method for the system that minimizes both per-core energy consumption and VR energy consumption using dynamic voltage frequency scaling and VR phase scaling by solving a comprehensive convex model. In order to improve the accuracy of the task latency model, a new task model considering both computation and memory access of the task is also developed. Furthermore, for better scalability and lower online overhead, we decompose our proposed convex method into two stages: 1) an offline stage and 2) an online stage. During the offline stage, we explore the convex model by assuming different numbers of active phases of the VR, various workload pressures and workload characteristics to collect the optimal frequency assignments under different scenarios. During the online stage, the specific frequency assignment for cores and optimal active phase number of the VR are selected and applied based on the actual workload pressure and its characteristics running on the cores. Experiments on real benchmarks show that when compared with the state-of-the-art approaches, which are oblivious to VR overheads and exploit slack time to achieve energy minimization, our method can achieve a significant energy saving of up to 22.4% with negligible online overhead. [ABSTRACT FROM AUTHOR]

Suppression of the higher-order azimuthal mode competition is significant in the design of an X-band triaxial klystron amplifier (TKA), since coaxial waveguide cannot cut off the low-order coaxial TE modes. In this article, the generation mechanism of the higher-order azimuthal mode competition is analyzed, and a slotted coaxial waveguide is specially designed to suppress the coaxial TE81 mode leakage. The results show that the reflection coefficient of the slotted coaxial waveguide can be optimized over 99%, and the reflection coefficient is sensitive to the variation of the slot length rather than the aperture angle or the thickness of the inner conductor. The combination of two identical slotted coaxial waveguides staggered azimuthally can effectively reflect the arbitrary polarized coaxial TE81 mode. Together with the reflectors specially designed to suppress the coaxial TE61 mode leakage, all of the asymmetric TE modes’ competitions are restrained in this device. In the 3-D particle-in-cell simulation, the pulse-shortening resulting from the coaxial TE81 mode leakage is successfully suppressed with the slotted coaxial waveguide. The output power of the TKA achieves 1.28 GW, and the efficiency and gain are 34.5% and 41.5 dB, respectively. The slotted coaxial waveguide together with the reflectors eliminates the utilization of attenuation materials. [ABSTRACT FROM AUTHOR]