Search

Sarcasm is a sophisticated speech act that is intended to express contempt or ridicule on social communities such as Twitter. In recent years, the prevalence of sarcasm on the social media has become highly disruptive to sentiment analysis systems due to not only its tendency of polarity flipping but also usage of figurative language. It is observed that sarcastic texts often convey a humorous effect. Thus, determining the humor in texts can be pertinent to the successful detection of sarcasm, and vice versa. However, current works always regard sarcasm detection and humor identification as separate tasks. In this paper, we argue that these tasks should be treated as a joint, collaborative, effort, considering the semantic connections between sarcasm and humor expressed in texts. Enlightened by the multi-task learning strategy, we present a joint architecture that settles two highly pertinent tasks, sarcasm detection and humor identification. As the basic of deep neural networks, we learn both tasks jointly exploring weight sharing to capture the task-specific features for each task and task-cross features between the two tasks. Extensive experiments on real-world datasets demonstrate that our presented model consistently enhances both sarcasm detection and humor identification tasks consistently with the help of the strong semantic relationships, achieving much better performance than state-of-the-art baselines.

Adopt robot arm for automatic spraying of footwear soles to solve the problems of low efficiency of manual spraying and substandard quality of spraying glue. Established a spray torch model with the pressure of the spraying swath $p_{f}$ , the atomizing pressure $p_{w}$ , spraying flow $q$ , spraying distance $h$ , and spraying angle $\theta $ as the variables, a visual trajectory offset function was established by introducing the sidewall $L$ of the sole and the normal vector $\overline n$ of the trajectory point, and the B-spline curve was used for interpolation to the offset trajectory. The improved non-dominated sorting genetic algorithm (INSGA-Ï) was adopted to solve the spraying trajectory with time, energy, and smoothness as the objectives. A normalized weight function was established that treated the three objectives with equal importance to optimize the solutions. Furthermore, a simulation to optimize the glue spray trajectory and glue spraying experiments were conducted on the sidewall of a shoe sole with a size of 41. As revealed in the results, provided that energy and smoothness are ensured, the proposed model can spray glue on the sidewall of the footwear sole uniformly within 7.27 s, no glue overflow and leakage phenomenon, and the spraying effect is uniform, peeling strength is more than 2.5Kgf/cm, greater than the process requirements of the index, and solved the bottleneck of shoe sole spraying.

To improve the performance of position sensorless control with a single current sensor in low-speed range, an improved current reconstruction method is proposed. High-frequency voltage is injected to estimate the rotor position. The time interval between two sampling moments is fixed as the minimum sampling window time. Therefore, the phase current can be compensated by an online adaptive one-step method. And the time-aligned three-phase currents can be reconstructed. The influence of the error between time-aligned reconstructed currents and the full sensors sampling currents on sensorless control is discussed. With the proposed method, the injection frequency can be improved to half the switching frequency. Finally, a permanent magnet synchronous motor is sensorless controlled with the proposed method in the experiments. [ABSTRACT FROM AUTHOR]

The grid-connected inverter may get unstable due to variation of grid impedance. To improve the system stability, an adaptive active damper, which emulates an adaptive tuning virtual resistor, can be introduced at the point of common coupling. However, subject to the control bandwidth, the virtual resistor is actually a frequency- dependent impedance, and its damping performance is impaired. In this article, we propose an impedance shaping method to make the virtual impedance be resistive in a wider frequency range, and it is implemented under hybrid frame for directly controlling the active power of the active damper and eliminating the influence of the cross-coupling terms on the emulation of a virtual resistor. Finally, the prototypes of a 3-kVA three-phase active damper and a 6-kVA three-phase LCL-type grid-connected inverter are fabricated and tested in the lab. The experimental results are provided to verify the effectiveness of the proposed control methods for the three-phase adaptive active damper. [ABSTRACT FROM AUTHOR]

In this article, in order to improve the bearing capacity per unit area and reduce the nonlinearity and couplings of three-pole magnetic bearing, which is driven by a three-phase power inverter, a three-degree-of-freedom (3-DOF) six-pole hybrid magnetic bearing (HMB) is proposed. To realize the larger bearing capacity and smaller volume, a multilevel design optimization is proposed to conduct multiobjective optimization. At first, the structure and working principle of the ac–dc 3-DOF six-pole HMB are introduced. Then, the suspension force mathematical models of the 3-DOF HMB based on the Maxwell tensor method are established. Next, the design variables and optimization objectives are selected, and the comprehensive sensitivity analysis is adopted to divide the design variables into three levels; the response surface method and multiobjective particle swarm optimization algorithm are applied to realize a compromise among the three optimization objectives. Finally, the results of the simulation and experiment verify the validity and correctness of the modeling and optimization design. [ABSTRACT FROM AUTHOR]

The general tensor-based methods can recover missing values of multidimensional images by exploiting the low-rankness on the pixel level. However, especially when considerable pixels of an image are missing, the low-rankness is not reliable on the pixel level, resulting in some details losing in their results, which hinders the performance of subsequent image applications (e.g., image recognition and segmentation). In this article, we suggest a novel multiscale feature (MSF) tensorization by exploiting the MSFs of multidimensional images, which not only helps to recover the missing values on a higher level, that is, the feature level but also benefits subsequent image applications. By exploiting the low-rankness of the resulting MSF tensor constructed by the new tensorization, we propose the convex and nonconvex MSF tensor train rank minimization (MSF-TT) to conjointly recover the MSF tensor and the corresponding original tensor in a unified framework. We develop the alternating directional method of multipliers (ADMMs) to solve the convex MSF-TT and the proximal alternating minimization (PAM) to solve the nonconvex MSF-TT. Moreover, we establish the theoretical guarantee of convergence for the PAM algorithm. Numerical examples of real-world multidimensional images show that the proposed MSF-TT outperforms other compared approaches in image recovery and the recovered MSF tensor can benefit the subsequent image recognition. [ABSTRACT FROM AUTHOR]

Traditional nonisolated multilevel inverters based on the buck–boost principle usually have some disadvantage such as necessary dead time or capacitor precharging. To overcome these, a new transformerless single-stage buck–boost voltage source inverter topology and its control method are proposed. In addition, the proposed topology has greater advantages in cost and loss. In this letter, the topological characteristics and working mode of the inverter are analyzed, and a modulation scheme that can ensure the balance of the output capacitor voltage is designed. Finally, a hardware experimental prototype is built, and the proposed scheme is experimentally verified based on DSP TMS320F28335 and FPGA XC6SLX9-2TQG144C digital controller. [ABSTRACT FROM AUTHOR]

This brief presents a broadband SiGe power amplifier that achieves 30.1 dBm peak $\text{P}_{SAT}$ , 38.6% peak power-added efficiency (PAE), 90.9% large-signal fractional bandwidth (FBW) and a 3 dB bandwidth from 5.4 to 13.8 GHz. A balanced topology with two-stage, four-way combing is utilized to achieve watt-level output power and high-power efficiency in a wide range of operating frequency. Power stage is optimized and simulated from the perspective of high output power. At the output stage, a 90° coupler and two transformers are co-optimized for broadband output power combining as well as power matching. This balanced PA is implemented in a 0.13- $\boldsymbol {\mu }\text{m}$ SiGe BiCMOS process with a chip area (with pads) of 2.31 mm2. [ABSTRACT FROM AUTHOR]

A simple yet effective approach for programmable gain amplifier (PGA) with a high precision decibel (dB)-linear characteristic and wide bandwidth is presented in this brief. This novel approach is presented by innovatively combining two techniques. Based on the binary-weighted switching technique, the circuit can approach an accurate pseudo-exponential function without using an additional exponential generator for gain control. By adopting the active feedback circuit structure, the bandwidth of PGA can be extended and it also contributes to improve the robustness, both of which cleverly reached a good consistency. The proposed PGA is fabricated in a $0.18~{\mu }\text{m}$ CMOS technology, the core circuit occupies an area of 0.11 mm2. Operated at a supply voltage of 1.8 V, the PGA consumes 8.21 mW. The measurement results show that the voltage gain of the proposed PGA can be controlled from −4.2 dB to 25.9 dB with a gain error less than ±0.09 dB, and the minimum −3dB bandwidth is almost 630 MHz. The input 1dB compression point is measured from −28.7 to 0.5 dBm, while the input referred noise is from 22.4 to 70.9 nV/ ${{\sqrt {}}}$ Hz. [ABSTRACT FROM AUTHOR]

This brief presents a 2.0~42.0 GHz ultra-wide bandwidth Cascode distributed low-noise amplifier (CDLNA) MMIC design. With the proposed coupled-line (CL) sections between each drain-artificial transmission line (D-ATML) node and related gate-artificial transmission line (G-ATML) node, the traveling signal is reused to enhance the gain and broaden the operation bandwidth simultaneously. To improve the gain further, the inductive gain peaking has also been employed at each drain of common-source (CS) and common-gate (CG) transistors. Meanwhile, an active termination technique incorporating active and passive elements has been adopted to achieve better noise figure (NF) performance. Employing the proposed techniques, an LNA prototype has been designed and fabricated in a 0.15- $\mu \text{m}$ GaAs E-Mode pHEMT technology with the chip size of 1.53 mm2 including all the pads. From test results, the proposed CDLNA demonstrates a peak gain of 14.1 dB, the best NF of 2.1 dB, a group delay of 102 ± 58.4 ps, and 23.25/14.7 dBm best OIP3/OP1dB, respectively. The total power consumption is 129 mW under 3 V VDD. [ABSTRACT FROM AUTHOR]

This article is concerned with the problem of compensation-based output feedback control for Takagi–Sugeno fuzzy Markov jump systems subject to packet losses. The phenomenon of packet losses is assumed to randomly occur in the feedback channel, which is modeled by a Bernoulli process. Employing the single exponential smoothing method as a compensation scheme, the missing measurements are predicted to help offset the impact of packet losses on system performance. Then, an asynchronous output feedback controller is designed by the hidden Markov model. Based on the mode-dependent Lyapunov function, some novel sufficient conditions on the controller existence are derived such that the closed-loop system is stochastically stable with strict dissipativity. Besides, an algorithm for determining the optimal smoothing parameter is proposed. Finally, the validity and advantages of the design approach are manifested by some simulation results. [ABSTRACT FROM AUTHOR]

The performance of decomposition-based algorithms is sensitive to the Pareto front shapes since their reference vectors preset in advance are not always adaptable to various problem characteristics with no a priori knowledge. For this issue, this article proposes an adaptive reference vector reinforcement learning (RVRL) approach to decomposition-based algorithms for industrial copper burdening optimization. The proposed approach involves two main operations, that is: 1) a reinforcement learning (RL) operation and 2) a reference point sampling operation. Given the fact that the states of reference vectors interact with the landscape environment (quite often), the RL operation treats the reference vector adaption process as an RL task, where each reference vector learns from the environmental feedback and selects optimal actions for gradually fitting the problem characteristics. Accordingly, the reference point sampling operation uses estimation-of-distribution learning models to sample new reference points. Finally, the resultant algorithm is applied to handle the proposed industrial copper burdening problem. For this problem, an adaptive penalty function and a soft constraint-based relaxing approach are used to handle complex constraints. Experimental results on both benchmark problems and real-world instances verify the competitiveness and effectiveness of the proposed algorithm. [ABSTRACT FROM AUTHOR]

In this article, the distributed fusion estimation issue with the dimensionality reduction strategy under DoS attacks and deception attacks is investigated for a class of stochastic uncertain systems with fading measurements. The stochastic uncertainties existed in the system and measurement equations are represented by state-dependent noises. The fading measurements are depicted by stochastic variables with known statistics. Then, a novel attack and compensation model is proposed to display the randomly occurring behaviors of the DoS attacks and the deception attacks within a unified framework. Furthermore, a distributed multisensor fusion estimation (DMSFE) algorithm is presented. An explicit form of dimensionality reduction is designed against attacks. Stability conditions are derived such that the mean square errors (MSEs) of the proposed DMSFE are bounded. A sequential covariance intersection fusion estimator (SCIFE) is designed to prevent the cross fusion covariance matrices calculating, which owns lower accuracy by smaller computation cost than DMSFE. An illustrative example is provided to show the effectiveness and merits of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Deep neural networks have achieved great performance on blind Image Quality Assessment (IQA), but it is still challenging for using one network to accurately predict the quality of images with different distortions. In this paper, a Distortion-Aware Convolutional Neural Network (DACNN) is proposed for blind IQA, which works effectively for not only synthetically distorted images but also authentically distorted images. The proposed DACNN consists of a distortion aware module, a distortion fusion module, and a quality prediction module. In the distortion aware module, a Siamese network-based pretraining strategy is proposed to design a synthetic distortion-aware network for full learning the synthetic distortions, and an authentic distortion-aware network is used for extracting the authentic distortions. To efficiently fuse the learned distortion features, and make the network pay more attention to the essential features, a weight-adaptive fusion network is proposed to adaptively adjust the weight of each distortion. Finally, the quality prediction module is adopted to map the fused features to a quality score. Extensive experiments on four authentic IQA databases and four synthetic IQA databases have proved the effectiveness of the proposed DACNN. [ABSTRACT FROM AUTHOR]

Nowadays, field programmable gate arrays (FPGAs) have been widely adopted to serve as accelerators in artificial intelligence and big data related applications. Since the static random access memory (SRAM)-based FPGA is suffering from limited density and high leakage power, nonvolatile FPGAs have been proposed, where SRAM is replaced with emerging nonvolatile memories (NVMs). Multilevel cell (MLC), which can store multiple bits within one memory cell, further improves the density of nonvolatile FPGAs and shows great potential to enable large on-chip memory. However, it suffers from limited lifetime. In this article, we propose a wear leveling scheme to improve lifetime of MLC-based nonvolatile FPGAs. Instead of generating a series of configuration files for runtime reconfiguration, we propose to identify write-heavy MLC regions and dynamically transform them to durable single-level cell (SLC) mode. Specifically, we propose three modules: 1) pertaining to write behavior monitor; 2) approximate cost calculator; and 3) mode transformation manager to achieve adaptive mode transformations. We consider FPGA features to design these modules, which is different from implementations for MLC-SLC transformation in CPU architecture. Evaluation shows that the proposed scheme can improve lifetime for MLC nonvolatile FPGAs by $6.03\times $ , at cost of 12.5% storage overhead. [ABSTRACT FROM AUTHOR]

Pole-zero cancellation technique has drawn much attention to tackle the filter resonance of the LCL-type grid-connected inverter for simplifying the stability-oriented design. However, its potential stability risk was rarely concerned. As one representative of this technique, the weighted average current (WAC) control is assessed. It is found that the LCL resonance in the state variable WAC is cancelled out by pole-zero cancellation, but the ones in some uncontrollable or unobservable state variables such as the grid current are still inherited, resulting in harmonic instability. From this perspective, this article appeals that this technique should be cautiously used or even avoided. An extended WAC control which avoids pole-zero cancellation is proposed accordingly. The theoretical expectations are finally validated by simulation and experimental results. [ABSTRACT FROM AUTHOR]

For multimodal representation learning, traditional black-box approaches often fall short of extracting interpretable multilayer hidden structures, which contribute to visualize the connections between different modalities at multiple semantic levels. To extract interpretable multimodal latent representations and visualize the hierarchial semantic relationships between different modalities, based on deep topic models, we develop a novel multimodal Poisson gamma belief network (mPGBN) that tightly couples the observations of different modalities via imposing sparse connections between their modality-specific hidden layers. To alleviate the time-consuming Gibbs sampler adopted by traditional topic models in the testing stage, we construct a Weibull-based variational inference network (encoder) to directly map the observations to their latent representations, and further combine it with the mPGBN (decoder), resulting in a novel multimodal Weibull variational autoencoder (MWVAE), which is fast in out-of-sample prediction and can handle large-scale multimodal datasets. Qualitative evaluations on bimodal data consisting of image-text pairs show that the developed MWVAE can successfully extract expressive multimodal latent representations for downstream tasks like missing modality imputation and multimodal retrieval. Further extensive quantitative results demonstrate that both MWVAE and its supervised extension sMWVAE achieve state-of-the-art performance on various multimodal benchmarks. [ABSTRACT FROM AUTHOR]

The real-time localization problem for a class of multiagent systems with imperfect channels under directed graphs is investigated in this article. A successive distributed localization estimation algorithm based on iterative learning is proposed to solve the localization problem of the agent with bounded communication delays. The barycentric coordinates are used to represent the relative position between agents. An estimator is proposed to inspect whether the received data packet is the latest through the iteration stamp on the packet. Meanwhile, an intermittent distributed localization algorithm is devised to handle with the random data loss of agents in real-time localization problem. The relationship between the convergence rate and the probability of the data loss is given in the sense of mathematical expectations. Both numerical simulation and Qbot-2e robot experiment are provided to verify the effectiveness and the feasibility of the aforementioned methods. [ABSTRACT FROM AUTHOR]

Cable fault location is generally much more difficult if cable defect has not reached the level of serious breakdown. Traditional methods such as time domain reflectometry (TDR) and partial discharge detection cannot solve this problem well due to their low sensitivity. This paper proposes a new non-destructive detection method based on the impedance spectroscopy (IS), which can effectively locate soft faults in the cable. In this paper, a numerical model for cable with distributed parameters is established firstly. Based on this model, cable faults are extracted from the imaginary part of impedance spectrum with Discrete Fourier Transform (DFT). The Chebyshev window is introduced to suppress interference and enhance signal-to-noise ratio, which improves the sensitivity and accuracy of fault location significantly. Experiments and simulations show that this method can well locate multi-segment defects in cables with good accuracy. The window function successfully amplifies the normalized amplitude of faults and improves the positioning accuracy. Compared with other method, this method reduces distortion points effectively. The fault location error is within 0.53%, which has great potential for field application. [ABSTRACT FROM AUTHOR]

This article proposes a unified framework to investigate the leader-following and leaderless consensus problem of general linear multiagent systems under the directed communication topology only containing a spanning tree. To further reduce the use of resources for the control objective, an energy-saving control algorithm is introduced composed of double dynamic event-triggered mechanisms that work independently: one intends to control the communication of agents with their neighbors and the other to decide the update of controllers. It is shown that the control algorithm performs well compared to most existing control algorithms in terms of the communication cost between agents and the update cost of controllers. In addition, a new procedure to choose parameters is obtained by the developed Lyapunov stabilty method, with the potential of achieving less conservative parameters than related results. Finally, the effectiveness of the proposed control scheme is verified by numerical simulations. [ABSTRACT FROM AUTHOR]

We propose a dual circularly polarized (CP) substrate integrated waveguide (SIW) cavity-backed antenna with wide axial ratio (AR) beamwidth, which can generate the right-hand CP (RHCP) or left-hand CP (LHCP) radiation through exciting different ports. Dual CP radiations are accomplished by an SIW directional coupler (DC) in the bottom layer, which provides the desired amplitude and phase conditions for two orthogonal slots. By superposing phase centers of linearly polarized (LP) electric field elements, the proposed antenna achieves a good performance of 3 dB AR beamwidth of 165° in the desired azimuth cut plane. Furthermore, an eight-element dual CP linear phased array with a wide-angle beam scanning ability is comprised by this antenna prototype. Numerical and measured results show that the array antenna can scan its main lobe from – 59° to 58°. In general, the proposed antenna array fed by the SIW structure satisfies the requirement of a multipolarization wireless communication system for low profile, simple structure, and easy integration with planar circuits. [ABSTRACT FROM AUTHOR]

This article considered the persistent real-time localization problem of dynamic multiagent systems with repetitive running characteristics under directed graph. The trajectory length of the agent is randomly varying caused by system constraints and external environment. A novel distributed iterative learning localization estimation method with full historical average data compensation is designed. The average value of all available historical data in the previous operations is used to compensate the incomplete trajectory. The designed diagnosis mechanism and search mechanism are used to determine whether the agent has stopped running and to screen out the available historical data. In order to reduce the computation burden, an improved distributed localization algorithm with receding horizon historical average data compensation is proposed. The asymptotic convergence of the estimation algorithms in the sense of mathematical expectation is derived through the rigorous analysis. Meanwhile, the influence of the incomplete repetitive trajectory of the agent on estimation error and convergence rate is also analyzed. The experimental result based on QBot-2e robot platform verifies the realizability of the proposed method. [ABSTRACT FROM AUTHOR]

A Ku-band electronic 2-dimensional (2-D) beam-scanning transmitarray employing a new reconfigurable dual-layer Huygens element is developed in this article. The Huygens element consists of two metallic crosses printed on two layers of a dielectric substrate, which enables a near nonreflection Huygens resonance. A 1 bit phase compensation with low transmission loss is realized by controlling two p-i-n diodes on the element. Compared with many other reconfigurable transmitarray elements using multilayer structures with metallic vias, the proposed reconfigurable Huygens element has a much simpler configuration with a simpler biasing network, and it is not affected by multilayer alignment errors. This particularly facilitates large aperture array development at higher frequencies. To validate the design concept, an electronically reconfigurable transmitarray with the proposed element is fabricated at 13 GHz. Good agreement between the measured and simulated results is found, showing 2-D scanning beams within ±50° in the E-plane and ±40° in the H-plane with a maximum realized gain of 18.4 dBi. [ABSTRACT FROM AUTHOR]

Natural Language Video Localization (NLVL) aims to locate a target moment from an untrimmed video that semantically corresponds to a text query. Existing approaches mainly solve the NLVL problem from the perspective of computer vision by formulating it as ranking, anchor, or regression tasks. These methods suffer from large performance degradation when localizing on long videos. In this work, we address the NLVL from a new perspective, i.e., span-based question answering (QA), by treating the input video as a text passage. We propose a video span localizing network (VSLNet), on top of the standard span-based QA framework (named VSLBase), to address NLVL. VSLNet tackles the differences between NLVL and span-based QA through a simple yet effective query-guided highlighting (QGH) strategy. QGH guides VSLNet to search for the matching video span within a highlighted region. To address the performance degradation on long videos, we further extend VSLNet to VSLNet-L by applying a multi-scale split-and-concatenation strategy. VSLNet-L first splits the untrimmed video into short clip segments; then, it predicts which clip segment contains the target moment and suppresses the importance of other segments. Finally, the clip segments are concatenated, with different confidences, to locate the target moment accurately. Extensive experiments on three benchmark datasets show that the proposed VSLNet and VSLNet-L outperform the state-of-the-art methods; VSLNet-L addresses the issue of performance degradation on long videos. Our study suggests that the span-based QA framework is an effective strategy to solve the NLVL problem. [ABSTRACT FROM AUTHOR]

Accurate and timely nowcasting precipitation has huge social and economic benefits. However, the changes in clouds including expansion, dissipation, and distortion are extremely complex, which exacerbates the difficulty of forecasting. Fortunately, it still follows certain meteorological laws, which can be explored based on spatiotemporal information but are not fully considered by previous models. In this article, we design two modules to focus on these messages based on atmospheric characteristics. Specifically, the spatial local attention memory (SLAM) module combines local attention and memory mechanism to capture the meteorological spatial relationship, while the time difference memory (TDM) module combines differential technology and memory mechanism to capture the meteorological temporal variants. We combine these two modules with PredRNN and propose PrecipLSTM to sufficiently capture the spatiotemporal dependencies of radar data. We do exhaustive experiments with five baselines on four radar datasets. It is verified that PrecipLSTM achieves state-of-the-art results with fewer parameters than the previous state-of-the-art method. [ABSTRACT FROM AUTHOR]

Text clustering is a critical step in text data analysis and has been extensively studied by the text mining community. Most existing text clustering algorithms are based on the bag-of-words model, which faces the high-dimensional and sparsity problems and ignores text structural and sequence information. Deep learning-based models such as convolutional neural networks and recurrent neural networks regard texts as sequences but lack supervised signals and explainable results. In this paper, we propose a deep feature-based text clustering (DFTC) framework that incorporates pretrained text encoders into text clustering tasks. This model, which is based on sequence representations, breaks the dependency on supervision. The experimental results show that our model outperforms classic text clustering algorithms and the state-of-the-art pretrained language model, i.e., BERT, on almost all the considered datasets. In addition, the explanation of the clustering results is significant for understanding the principles of the deep learning approach. Our proposed clustering framework includes an explanation module that can help users understand the meaning and quality of the clustering results. [ABSTRACT FROM AUTHOR]

This article considers the quasisynchronization of memristive neural networks (MNNs) with communication delays via event-triggered impulsive control (ETIC). In view of the limited communication and bandwidth, we adopt a novel switching event-triggered mechanism (ETM) that not only decreases the times of controller update and the amount of data sent out but also eliminates the Zeno behavior. By using an appropriate Lyapunov function, several algebraic conditions are given for quasisynchronization of MNNs with communication delays. More important, there is no restriction on the derivation of the Lyapunov function, even if it is an increasing function over a period of time. Then, we further propose a switching ETM depending on communication delays and aperiodic sampling, which is more economical and practical and can directly avoid Zeno behavior. Finally, two simulations are presented to validate the effectiveness of the proposed results. [ABSTRACT FROM AUTHOR]

Capacitor-current active damping has been widely used in voltage source inverters with LC or LCL filters. However, the time delay poses an adverse impact on the damping performance, causing a negative-resistance behavior, and thus constraints the system stability. To address this issue, this letter proposes a compensation method, which eliminates the adverse impact by removing the time delay out of the capacitor-current loop. Consequently, the damping performance behaves as a constant positive resistance, and thereby the stable region of the damping gain is notably enlarged. Experimental results verify the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Increases in speed and sensitivity enabled rapid clinical adoption of optical coherence tomography (OCT) in ophthalmology. Recently, visible-light OCT (vis-OCT) achieved ultrahigh axial resolution, improved tissue contrast, and provided new functional imaging capabilities, demonstrating the potential to improve clinical care further. However, limited speed and sensitivity caused by the high relative intensity noise (RIN) in supercontinuum lasers impeded the clinical adoption of vis-OCT. To overcome these limitations, we developed balanced-detection vis-OCT (BD-vis-OCT), which uses two calibrated spectrometers to cancel RIN and other noises. We analyzed the RIN to achieve robust subpixel calibration between the two spectrometers and showed that BD-vis-OCT reduced the A-line noise floor by up to 20.5 dB. Metrics comparing signal-to-noise-ratios showed similar image qualities across multiple reference arm powers, a hallmark of operation near the shot-noise limit. We imaged healthy human retinas at an A-line rate of 125 kHz and a field-of-view up to 10 mm $\times4$ mm. We found that BD-vis-OCT revealed retinal anatomical features previously obscured by the noise floor. [ABSTRACT FROM AUTHOR]

In this article, we propose using new machine learning (ML)-based optimization methods as an alternative to traditional optimization methods, for complex antenna designs. This is an efficient methodology to tackle computational challenges, as it is capable of handling a large number of design parameters and is more efficient as well as informative. The proposed technique is applied for modeling the gain performance in the principal plane of a monopole antenna when its radiation properties are modified by placing spatially dependent dielectric material around it. Using the proposed methodology, the dielectric constant values are mapped to the gain pattern of the design. We use two ML techniques for this purpose, namely Gaussian process (GP) regression and artificial neural network (ANN). Once each of these models is obtained, they are further used for estimating the dielectric constant values that can suggest optimal directions to modify gain patterns for single-beam and multiple-beam patterns rather than the conventional omnidirectional pattern of a monopole antenna. The performance of this technique is compared with heuristic optimization techniques, such as genetic algorithms (GAs). The proposed method proves to be quite accurate in spite of being a high-dimensional nonlinear problem. A prototype of a monopole design with three-beam gain pattern is fabricated and tested. The measurement results agree well with the simulation results. The proposed methodology can provide useful and scalable optimization tools for computationally intensive antenna design problems. [ABSTRACT FROM AUTHOR]

In this brief, the dynamic event-triggered control is investigated for a novel class of singularity-perturbed dynamic networks (SPDNs) with multi-time-scale. First, for a single agent, the two-time-scale characteristic described by a small singularity-perturbed parameter (SPP) is considered. Then, every agent is equipped with a personalized SPP $\varepsilon _{i}$ , which denotes the multi-time-scale characteristic of the SPDNs. Owing to the SPP $\varepsilon _{i}$ , the restriction in most the existing works that every agent must have the same SPP can be removed. Further, the networks with single-time-scale, two-time-scale, or multi-time-scale can all be modeled by the proposed SPDNs by selecting an appropriate value of $\varepsilon _{i}$. A state feedback controller based only on the slow dynamics information is considered. Besides, a dynamic event-triggered mechanism (ETM), where the threshold parameters are adjusted as the slow state fluctuations, is introduced to reduce the communication frequency reasonably. By means of Lyapunov stability theorem and a decoupled method, some stability conditions that are independent of the scale of the SPDNs are obtained. Finally, the effectiveness and advantages of the proposed method over existing ones are demonstrated through an electronic circuit system. [ABSTRACT FROM AUTHOR]

Deep probabilistic aspect models are widely utilized in document analysis to extract the semantic information and obtain descriptive topics. However, there are two problems that may affect their applications. One is that common words shared among all documents with low representational meaning may reduce the representation ability of learned topics. The other is introducing supervision information to hierarchical topic models to fully utilize the side information of documents that is difficult. To address these problems, in this article, we first propose deep diverse latent Dirichlet allocation (DDLDA), a deep hierarchical topic model that can yield more meaningful semantic topics with less common and meaningless words by introducing shared topics. Moreover, we develop a variational inference network for DDLDA, which helps us to further generalize DDLDA to a supervised deep topic model called max-margin DDLDA (mmDDLDA) by employing max-margin principle as the classification criterion. Compared to DDLDA, mmDDLDA can discover more discriminative topical representations. In addition, a continual hybrid method with stochastic-gradient MCMC and variational inference is put forward for deep latent Dirichlet allocation (DLDA)-based models to make them more practical in real-world applications. The experimental results demonstrate that DDLDA and mmDDLDA are more efficient than existing unsupervised and supervised topic models in discovering highly discriminative topic representations and achieving higher classification accuracy. Meanwhile, DLDA and our proposed models trained by the proposed continual learning approach cannot only show good performance on preventing catastrophic forgetting but also fit the evolving new tasks well. [ABSTRACT FROM AUTHOR]

Salt body extraction plays an important role in the analysis of salt structures and the exploration of oil and gas. Seismic attributes and edge detection algorithms, which require manual effort, are the conventional methods of extracting salt boundaries from seismic images. Convolutional neural networks (CNNs) have become the state-of-the-art automatic segmentation method for seismic interpretation. However, the fully automatic results of the extraction of salt boundaries may still need to be modified to become accurate and robust enough for practical production. We present a novel deep-learning-based interactive segmentation method for extracting salt boundaries. To incorporate the interaction points into our method, we transform positive and negative points into two Euclidean distance maps (EDMs), which are combined with seismic images to train our CNN model. The model is composed of a U-net and a pyramid pooling module (PPM), and it is trained on the Tomlinson Geophysical Services (TGS) Salt Identification Challenge dataset. Then, we use a graph cut algorithm to refine the likelihood maps predicted by our CNN model and, subsequently, update the salt boundaries. Some field examples show that the proposed method outperforms fully automatic CNN methods with a higher matching degree of the ground truth. [ABSTRACT FROM AUTHOR]

Multiple-port dc–dc converters are characterized by a variety of kinds and a large number of circuit topologies. In this article, we aim to reveal the intrinsic relationship among the topologies of multiple-port dc–dc converters and propose the topology derivation method. First, the voltage- and current-type ports are summarized from basic dc–dc converters, and the approach of converting current-type ports into voltage-type ports is discussed. Then, according to Kirchhoff's law, four types of multiple-input multiple-output converters named input-port-series output-port-series, input-port-series output-port-parallel, input-port-parallel output-port-series, and input-port-parallel output-port-parallel are presented. Second, the topology derivation method of multiple-port bidirectional dc–dc converters based on the voltage-type ports is studied in terms of power flow paths in various operation modes, and then the topology optimization method is proposed. Particularly, a flow diagram for the optimal design procedure is given to guide the topology derivation for some specific requirements. Based on the proposed approach, a family of multiple-port dc–dc converters can be derived, which provides lots of viable candidates for practical engineering. Furthermore, one derived converter named the parallel-type three-port bidirectional buck converter is analyzed in three operation modes to demonstrate the topology derivation. Finally, the effectiveness of the above theoretical analysis is verified by those experimental results. [ABSTRACT FROM AUTHOR]

In this article, the problem of a nonlinear system states estimation with multisensor stochastic scheduling is investigated. In order to solve the non-Gaussian property induced by the nonlinear transformation, the unscented transformation (UT) technique is applied. Since the sensor networks channel is limited, the stochastic event-triggered mechanisms (SETMs) are proposed to reduce the network transmission burden. Under the SETMs, the modified unscented Kalman filter is proposed. Additionally, the sufficient conditions are given to guarantee the stabilities of the error covariance and the estimation error. Finally, extensive examples are carried out. Performances evaluation and comparison with existing methods are given to demonstrate the superiority of the proposed methods. [ABSTRACT FROM AUTHOR]

In this article, the consensus problem of linear multiagent systems (MASs) is investigated, where an event-triggered mechanism combined with sampler is introduced to rationally utilize the network bandwidth and the communication energy. Different from most existing results on the event-triggered consensus of MASs, a time-varying communication delay (CD) with a more general form is constructed in this article. With the help of a new class of Lyapunov functional and advanced inequalities, some novel criteria for achieving the asymptotic consensus of the considered MASs under different CD cases are obtained. For demonstrating the superiority of the proposed method from the conservatism and the practicability, a compared example and an application of spacecraft formation flying are proposed, respectively. [ABSTRACT FROM AUTHOR]

In this brief, a variable-precision approximate floating-point multiplier is proposed for energy efficient deep learning computation. The proposed architecture supports approximate multiplication with BFloat16 format. As the input and output activations of deep learning models usually follow normal distribution, inspired by the posit format, for numbers with different values, different precisions can be applied to represent them. In the proposed architecture, posit encoding is used to change the level of approximation, and the precision of the computation is controlled by the value of product exponent. For large exponent, smaller precision multiplication is applied to mantissa and for small exponent, higher precision computation is applied. Truncation is used as approximate method in the proposed design while the number of bit positions to be truncated is controlled by the values of the product exponent. The proposed design can achieve 19% area reduction and 42% power reduction compared to the normal BFloat16 multiplier. When applying the proposed multiplier in deep learning computation, almost the same accuracy as that of normal BFloat16 multiplier can be achieved. [ABSTRACT FROM AUTHOR]

Digital control has been widely adopted in grid-connected inverters, thanks to easy implementation of complicated control algorithms. However, the digital control delay deteriorates the performance of the inverter and tends to cause system instability. In this article, the generating mechanism of the digital control delay, including computation delay and pulsewidth modulation delay, is analyzed. Then, a generalized real-time computation method with a dual-sampling mode is derived to directly eliminate the computation delay, based on two- and multilevel phase-legs. On this basis, the operation principle of the method for three- and single-phase inverters with two- and multilevel topologies is elaborated. Finally, the experimental and simulation results from three- and single-phase LCL-type grid-connected inverters are provided to verify the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

In this article, the output synchronization problem of passive multiagent systems (MASs) with transmission delays and switching graphs is addressed by a novel logic-based distributed switching mechanism. Our result shows that synchronization is reached for arbitrarily large and bounded constant, time varying, or distributed delays, which, compared with the existing results for passive MASs, has an obvious advantage. This delay robustness holds under the very weak connectivity assumptions on the underlying graph, that is, as long as the graph is uniformly jointly strongly connected and switches with a dwell time. The proposed algorithm is applied to the position synchronization problem of multiple robotic manipulators to show its applicability. [ABSTRACT FROM AUTHOR]

This article addresses the aperiodic sampled-data control problem for flexible spacecraft with stochastic actuator failures. Flexible spacecraft dynamics are approximated by a group of T–S fuzzy models due to strong nonlinearity, and the multi-stochastic failures of spacecraft are depicted by a time-continuous and state-discrete Markov chain. To reduce the design conservativeness, a membership-sampling-dependent Lyapunov–Krasovskii functional (MSDLKF) is introduced to utilize the information of fuzzy membership functions and aperiodic sampling modes. Furthermore, a number of reliable fuzzy controllers are designed to obtain the exponential attitude stabilization under the circumstances of stochastic failures. At the same time, disturbance attenuation is ensured. The solution of the fuzzy controller gains can be obtained by solving a set of linear matrix inequalities (LMIs). In the end, an example of the practical flexible spacecraft system is given to illustrate the feasibility and validity of the proposed fuzzy control methods. [ABSTRACT FROM AUTHOR]

This article addresses two important issues of causal inference in the high-dimensional situation. One is how to reduce redundant conditional independence (CI) tests, which heavily impact the efficiency and accuracy of existing constraint-based methods. Another is how to construct the true causal graph from a set of Markov equivalence classes returned by these methods. For the first issue, we design a recursive decomposition approach where the original data (a set of variables) are first decomposed into two small subsets, each of which is then recursively decomposed into two smaller subsets until none of these subsets can be decomposed further. Redundant CI tests can be reduced by inferring causalities from these subsets. The advantage of this decomposition scheme lies in two aspects: 1) it requires only low-order CI tests and 2) it does not violate $d$ -separation. The complete causality can be reconstructed by merging all the partial results of the subsets. For the second issue, we employ regression-based CI tests to check CIs in linear non-Gaussian additive noise cases, which can identify more causal directions by $x - E(x|Z) \_{}\!\!\_{}\!\!\!\!\!\shortparallel z$ (or $y - E(y|Z) \_{}\!\!\_{}\!\!\!\!\!\shortparallel z$). Consequently, causal direction learning is no longer limited by the number of returned $V$ -structures and consistent propagation. Extensive experiments show that the proposed method can not only substantially reduce redundant CI tests but also effectively distinguish the equivalence classes. [ABSTRACT FROM AUTHOR]