Search

When the training data and the test data do not obey the same distribution, the performances of many object detection methods always decrease greatly. Naturally, domain adaptation methods at feature level are proposed. The basic idea is to adapt the feature extraction network such that the feature distributions of the source and target domains match. We propose a new method that is built directly on the Faster R-CNN model, which not only aligns the source and target data features, but also forces their generated features closer together to further align the source and target domains. Moreover, compared with previous approaches, we construct a more powerful discriminator and a simple generator to solve the domain adaptation problem. The model works like a pair of scissors, so we call it Scissors Networks (SN). We conduct extensive experiments on popular datasets, including Cityscapes, Foggy Cityscapes, SIM10k and KITTI. The experimental results demonstrate that our algorithm is superior to the state-of-the-art deep learning based domain adaptation approaches.

Feature selection is a widely used dimension reduction technique to select feature subsets because of its interpretability. Many methods have been proposed and achieved good results, in which the relationships between adjacent data points are mainly concerned. But the possible associations between data pairs that are not adjacent are always neglected. Different from previous methods, we propose a novel and very simple approach for unsupervised feature selection, named MMFS (Multi-step Markov Probability Relationship for Feature Selection). The idea is using multi-step Markov transition probability to describe the relation between any data pair. Two ways from the positive and negative viewpoints are employed respectively to keep the data structure after feature selection. From the positive viewpoint, the maximum transition probability that can be reached in a certain number of steps is used to describe the relation between two points. Then, the features which can keep the compact data structure are selected. From the viewpoint of negative, the minimum transition probability that can be reached in a certain number of steps is used to describe the relation between two points. On the contrary, the features that least maintain the loose data structure are selected. The two ways can also be combined. Thus three algorithms are proposed. Our main contributions are a novel feature section approach which uses multi-step transition probability to characterize the data structure, and three algorithms proposed from the positive and negative aspects for keeping data structure and select the features to preserve such structure. The performance of our approach is compared with the state-of-the-art methods on eight real-world data sets, and the experimental results show that the proposed MMFS is effective in unsupervised feature selection.

Financial data forecasting is conducive to get a better understanding of the future economic situation. Recently, variational mode decomposition (VMD) is introduced into the field of financial data forecasting. However, the prediction accuracy of the current methods is low. We propose a new VMD-based financial data prediction model. In this model, the genetic algorithm is utilized to optimize the parameters of VMD. Then VMD decomposes the data sequence into long-term and short-term trends. Finally, we employ the long short-term memory (LSTM) network to predict the future data with inputs generated by VMD. The contributions are: (1) We propose an improved VMD and LSTM based financial data forecasting model; (2) A guideline on the parameter selection of VMD to process financial data is designed; (3) A prediction-error reducing method which reduces the inherent error caused by VMD insensitivity to fluctuation is proposed. Experimental results indicate our model is accuracy-promising and superior to the baseline models in one-step-ahead forecasting of financial time series.

Particle velocity (Vp) is investigated comprehensively in a two-dimensional circulating fluidized bed (CFB) riser with a superficial gas velocity (Ug) of 5–7 m/s and a solids circulation rate (Gs) ...

Silicate micro- and nano-additives are multifunctional in relation to cement systems. Their application can solve a wide range of technological problems while maintaining the economic efficiency of technical solutions. The effect of silicate additives and fillers is determined by their level of dispersion, due to which the technologies for producing nano- and submicro-sized dispersed materials are being developed. The combination of mechanochemical synthesis of modified calcium hydrosilicates with subsequent thermolysis makes it possible to produce calcium silicate dispersions (SCD), which differ in polymodality of the fractional composition including submicro (10–7–10–6 m) and microdimensional (≥10–6 m) modes. The main element of the technology is the use of modifying carbohydrate, which acts as a stabilizer of hydrated phases of silicates. A comparative study of SCD produced using sucrose (sSCD) and lactose (lSCD) revealed the effect of these carbohydrates on the properties of sSCD and lSCD, as well as their effectiveness as a component of cementitious composite binder. It was found that the level of adsorption of modifying carbohydrate determines the physical properties of SCD (granulometry, specific surface area). The relatively high residual content of free sucrose (0.24%) in the composition of sSCD prevents the consolidation of silicates nanoparticles formed during the thermolysis, causes a high content of submicro sized fractions and a high specific surface area with sSCD (26.3 ± 0.7 m2/g). Lactose is absorbed by the silicate phase; the residual content of free lactose does not exceed 0.028% of lSCD. The low content of stabilizing carbohydrate contributes to the development of nanoparticle consolidation, a decrease in the specific surface area of lSCD to 13.0 ± 0.2 m2/g and content of submicrosized fractions. The residual content of free carbohydrates and particle size characteristics of sSCD and lSCD determine the nature of their influence on Cement-SCD-based concrete setting and hardening. The presence of residual sucrose in the composition of sSCD and fine fractions determines the competitive nature of the processes of retardation of hardening and acceleration of hardening of the cement system due to the nucleation effect, as a result of which the curve of the setting time is extreme. In addition, the inhibitory effect of sucrose reduces the strength of concrete on the 7th day. By the 28th day, the inhibitory effect of sucrose has been overcome, and concrete samples demonstrate an 18% increase in compressive strength with a sSCD content of 30%. The low content of residual free lactose in the composition of lSCD causes the nucleation effect. As a result, there is a monotonous reduction in the setting time of concrete mix with an increase in the content of lSCD in the composition of HF, as well as a significant increase in concrete strength (up to 127%) on the 7th day. At the same time, on the 28th day the strength of concrete increases slightly

Removing rain streaks from single image is mathematically ill-defined and the known strategies tend to over-smooth the background and remove image details. To tackle this problem, this paper proposes a novel two-stage system. Firstly, in view of the frequency and direction of rain streaks, Gabor filter is first used to suppress rain streaks. Based on the bases extracted from natural images, several Gabor filtered images extracted from the input rain image are fused by a new proposed method, and the fused result provides a texture constraint for the non-rain part. Secondly, a low-rank model is adopted to describe rain streaks as they take on similar and repeated patterns in an imaging scene. By combining the texture constraint and the low-rank model, a global cost function is constructed. Finally, an optimization strategy is proposed based on the split Bregman technique, which solves the function with high performance. A large number of experimental results show the competitive performance and efficiency of the proposed method in comparison with the state-of-the-art methods, making it more applicable in real practices.

This article presents investigation of 3-D electrical capacitance tomography (ECT) sensors with different excitation strategies, including dual-electrode excitation in the same plane, in adjacent plane, in planes separated by one plane or two planes, and in the adjacent position or opposite position in the same or different planes. The results show that a 3-D ECT sensor with dual-electrode excitation in the opposite position provides higher mean capacitance and lower standard deviation than in the adjacent position. Dual-electrode excitation in different planes can increase the capacitance signal and decrease the dynamic range of capacitance measurements, and thus, decrease random errors and improve the quality of 3-D reconstructed images. A 3-D ECT sensor with dual-electrode excitation in planes separated by one plane and opposite position can provide small measurement uncertainty and high image quality.

An imaging system is designed with a liquid crystal (LC) lens as the focusing element. The aperture of the LC lens plays the role of the aperture stop of the system. An entrance pupil larger than t...

Most zero-shot learning (ZSL) methods aim to learn a mapping from visual feature space to semantic feature space or from both visual and semantic feature spaces to a common joint space and align them. However, in these methods the visual and semantic information are not utilized sufficiently and the useless information is not excluded. Moreover, there exists a strong bias problem that the instances from unseen classes always tend to be predicted as some seen classes in most ZSL methods. In this paper, combining the advantages of generative adversarial networks (GANs), a method based on bidirectional projections between the visual and semantic feature spaces is proposed. GANs are used to perform bidirectional generations and alignments between the visual and semantic features. In addition, cycle mapping structure ensures that the important information are kept in the alignments. Furthermore, in order to better solve the bias problem, pseudo-labels are generated for unseen instances and the model is adjusted along with them iteratively. We conduct extensive experiments at traditional ZSL and generalized ZSL settings, respectively. Experiment results confirm that our method achieves the state-of-the-art performances on the popular datasets AWA2, aPY and SUN.

Summary Stochastic gradient Markov chain Monte Carlo algorithms have received much attention in Bayesian computing for big data problems, but they are only applicable to a small class of problems for which the parameter space has a fixed dimension and the log-posterior density is differentiable with respect to the parameters. This paper proposes an extended stochastic gradient Markov chain Monte Carlo algorithm which, by introducing appropriate latent variables, can be applied to more general large-scale Bayesian computing problems, such as those involving dimension jumping and missing data. Numerical studies show that the proposed algorithm is highly scalable and much more efficient than traditional Markov chain Monte Carlo algorithms.

In industrial catalytic processes, coke deposition can cause catalyst deactivation by covering acid sites and/or blocking pores. The regeneration of deactivated catalysts, thereby removing the coke and simultaneously restoring the catalytic activity, is highly desired. Despite various chemical reactions and methods are available to remove coke, developing reliable, efficient, and economic regeneration methods for catalytic processes still remains a challenge in industrial practice. In this paper, the current progress of regeneration methods such as oxidation (air, ozone and oxynitride), gasification (carbon dioxide and water steam), and hydrogenation (hydrogen) is reviewed, which hopefully can shed some light on the design and optimization of catalysts and the related processes.

In this brief, a 32-channel analog front end circuit featured with high gain, high bandwidth, and low noise characteristics is designed. The proposed three-stage analog front end (TS-AFE) includes a variable gain transimpedance amplifier, a differential voltage amplifier, and an analog equalizer. The first stage adopts the shunt-feedback topology to optimize the overall noise, the second stage helps to enlarge the input dynamic range further, and the last stage boosts the channel bandwidth. Furthermore, a 6-bit current digital-to-analog converter (DAC) is integrated in the equalizer to produce a built-in threshold voltage. The 32-channel TS-AFE chip is fabricated with 0.18- ${\mu }\text{m}$ CMOS technology. The measurement results show that with the highest gain configuration, the circuit realizes 101.6 dB transimpedance gain, 190 MHz bandwidth, and 2.6 pA/sqrt(Hz) input noise current spectral density, respectively. The minimum detectable current achieves 0.4 ${\mu }\text{A}$ with SNR = 10. The maximum output voltage swing is greater than 1000 mV, and the linear dynamic range reaches 64 dB. The power consumption per channel is as low as 37 mW from a single 3.3 V supply.

With the increasing concern on the preservation of personal privacy, privacy-preserving data mining has become a hot topic in recent years. Spectral clustering is one of the most widely used clustering algorithm for exploratory data analysis and usually has to deal with sensitive data sets. How to conduct privacy-preserving spectral clustering is an urgent problem to be solved. In this study, the authors focus on introducing the notion of differential privacy, which is considered as the de facto standard of privacy-preserving data analysis, into spectral clustering. Specifically, by combining the well-studied constrained spectral clustering with the Wishart mechanism in a novel way, the authors propose a differentially private constrained spectral clustering (DP-CSC) algorithm. The DP-CSC algorithm is proved to capture asymptotic property and achieves ϵ -differential privacy. To illustrate the effectiveness and efficiency of DP-CSC, the authors conduct experiments on five real-word data sets. The results indicate that the DP-CSC algorithm can provide acceptable clustering accuracy with short running time while preserving individual privacy.

Occluded pedestrian detection has been a research difficulty in computer vision for a long time. The conventional approach to solve this problem is to learn partial detectors, which can be properly integrated for occlusion handling. However, the efficiency of this type of methods is limited in practical applications since the partial detectors can not cover all occlusion patterns. In this paper, an attention guided neural network model (AGNN) is proposed for the occlusion handling of pedestrian detections, which is inspired by the approaches of sentiment classification. Firstly, a fixed-size window slides on a still image without overlapping to generate a set of sub-images. Secondly, a convolutional neural network is employed to extract the high-level features from resulting sub-images. Then, the attention network performs local feature weighting, from which the features representing the body parts of pedestrians are selected. Finally, the feature sequences are classified by recurrent neural network in proper order based on the weighted results. In addition, we explore different mechanisms of attention guidance on the detector for the detections. Compared with the state-of-the-art methods on two standard pedestrian datasets, experimental results demonstrate the comparable performance of our approach in terms of miss rates.

A water resource allocation model coupling quantity and quality is a topic of concern in China. The water resources supply modelling is set up according to the water quality, and water quality has been emulated under the program of water resources allocation. It analysed the effect to water resources allocation which was caused by water quality and researched the technology of pollutants reduction under the protection target of water quality in water functional zone. Such a methodology can be used by basin managers in decision making and provides better performance in terms of both water resources and the environment. The methodology of water resource allocation model coupling quantity and quality is applied to the Huntai River Basin in China, under different scenarios. The results show that it is possible to reduce future water scarcity problems and improve the water environment. In addition, an optimal water resource allocation model can be useful for water saving and protection.

This brief presents a low-complexity hybrid readout architecture that can extract both timing and amplitude information of the return pulse concurrently for a light detection and ranging radar (Lidar) receiver. To reconstruct the short return pulse, the core circuit of one sampling and storage array with an embedded time-to-digital converter (SSA-TDC) is proposed. Instead of relying on the conventional power-hungry high-speed ADC, it selectively samples the interested return pulse with high speed and stores the voltages for later quantization in the long idle interval. In the preceding stage of the SSA-TDC, the analog front-end circuit cascaded of a narrow-bandwidth transimpedance amplifier, a voltage amplifier, and an equalizer circuit is included. The prototype chip of the eight readout channels is designed and fabricated with the 0.18- $\mu \text{m}$ CMOS process. The active circuit occupies an area of $1200\,\,\mu \text{m}\,\,\times 2100\,\,\mu \text{m}$ , and the power consumption of one single channel is 45 mW with 3.3-V supply. The proposed SSA-TDC has achieved a dynamic error of 180 ps in the experimental study.

China will launch in the forthcoming years a sample return mission called ZhengHe, to asteroid 469219 Kamo‘oalewa (provisional designation 2016HO3) and comet 133P/Elst-Pizarro. The mission will consist of an orbiter and a nano-lander. One of ZhengHe’s investigations is the radio science experiment, whose main objective is the asteroid GM estimate. In this paper, we conduct full numerical simulations of the radio science experiment using the wudogs software package, developed by Wuhan University. In addition to two-way Doppler measurements, we also include one-way on-board distance measurements. A list of parameters including the spacecraft initial conditions and the global asteroid GM are solved using a weighted least-squares fit. The simulation results indicate that the GM solution is very sensitive to the ephemeris error. We need an accuracy within 2 km on the ephemeris of the asteroid to achieve a reliable estimate of GM.

A robust adaptive chattering-free sliding mode (ACFSM) control method for electronic throttle (ET) system is proposed in this paper. It is well known that nonlinearities in the throttle system including friction, return-spring, limp-home (LH) and gear backlash affect the control accuracy of the throttle valve. Compared with the traditional sliding mode control methods, the ACFSM control not only overcomes the influence of nonlinearities and parameter uncertainties in the ET system, but also eliminates chattering in nature such that excellent throttle tracking performance and robustness are maintained. The proposed ACFSM control method is superior to the traditional sliding mode control in the following two aspects: 1) The chattering-free sliding mode control method is able to attenuate the control chattering without weakening the output tracking performance and robustness. 2) The upper bounds of the uncertainty and disturbance are not required any more in control design. They are online estimated by the adaptive law in the sense of Lyapunov. Moreover, due to the difficulty in selecting proper control parameters, the genetic algorithm (GA) is introduced to optimate the control parameters for the ACFSM controller prior to practical implementation. The comparative experimental results are given to demonstrate the excellent control performance of the proposed ACFSM control.

In recent years, the class imbalance problem that aims to correctly classify imbalanced data sets and improve the classification performance of minority instances has received attention. Such problem can be roughly described as one of the class(es) termed as minority class(es) contains much smaller instances than the others, also referred as majority class(es). To address this problem, a novel classification method called focused online random forest based on synthetic minority oversampling technique (FORF-SMOTE) is proposed in this paper and simply expressed as FORF-S, which constructs two online random forests respectively trained by original training dataset and new generated dataset, then further jointly consititute the model. Instead of oversampling the minority instances in data level, the algorithm in this paper is motivated by making the sampling strategies integrated into algorithm level to create classifiers, which can better identify the minority class. Moreover, the method is also compared with other state-of-the-art methods and the results have demonstrated that the proposed algorithm takes advantages of the aforementioned methods.

Multiferroic materials with multifunctional characteristics play a critical role in the field of microelectronics. In a perovskite oxide, ferroelectric polarization and ferromagnetism usually cannot coexist in a single-phase material at the same time. In this work, we design a superlattice structure composed of alternating BiFeO3 and BiMnO3 layers and illustrate how tuning the supercell size of epitaxial BiFeO3/BiMnO3 superlattices facilitates ferroelectric polarization while maintaining relatively strong ferromagnetism. A comprehensive investigation reveals that the enhanced ferroelectric polarization of BiMnO3 layers originates from the induction effect induced by a strong polarization field generated by the adjacent ferroelectric BiFeO3 layers. For the magnetic behavior, we consider the existence of interfacial antiferromagnetic superexchange interaction of Fe-O-Mn between BiFeO3 and BiMnO3 layers in our superlattices. This modulation effect of artificial superlattices provides a platform to accurately control the multiple order parameters in a multiferroic oxide system.

Porous particles are widely used in liquid–solid and gas-liquid-solid reactors, and accurate prediction of drag coefficients ( C D ) for porous particles is essential for designing and optimizing these reactors. The drag coefficients for porous particles, however, have been mainly obtained via theoretical analysis and numerical simulations. Only three experimental studies were carried out to study the drag force on porous spherical particles due to the difficulties in the preparation of model porous spheres with a homogeneous porosity and homogeneous surface properties. In this paper, a method to prepare model porous spheres which have a homogeneous porosity and homogeneous surface properties with a good sphericity for a wider range of relative permeability β has been presented. The details of this method were documented, and it showed that the porous spheres constructed could withstand rigid collisions at the maximum terminal velocity of 1 m/s and keep the structure intact with a negligible quantity of binders. Then gravity-driven settling experiments of a single porous sphere by use of particle tracking velocimetry (PTV) method were carried out to evaluate the feasibility of the use of the prepared porous spheres. Experimental results for 12 porous spheres, with particle Reynolds number ( R e ) ranging from 1 to 108 and relative permeability β from 21.6 to 315.6, agreed well with the previous results in the literature. The porous particles prepared by our proposed method can be potentially used for measuring drag coefficients of the porous permeable spheres in a wider range of particle Reynolds number and permeability and studying the complex interaction between the porous particles and the surrounding fluid.

Person re-identification (re-id) usually refers to matching people across disjoint camera views. Many existing methods focus on extracting discriminative features or learning distance metrics to make the intraclass distance smaller than interclass distances. These methods subconsciously assume that pedestrian images are well aligned. However, one major challenge in person re-id is the unconstrained spatial misalignment between image pairs due to view angle changes and pedestrian pose variations. To address this problem, in this paper, we propose Recurrent Matching Network of Spatial Alignment Learning (RMN-SAL) to simulate the human vision perception. Reinforcement learning is introduced to locate attention regions, since it provides a flexible learning strategy for sequential decision-making. A linear mapping is employed to convert the environment state into spatial constraint, comprising spatial alignment into feature learning. And recurrent models are used to extract information from a sequence of corresponding regions. Finally, person re-id is performed based on the global features and the features from the learned alignment regions. Our contributions are: 1) the recurrent matching network, which can subtly combine local feature learning and sequential spatial correspondence learning into an end-to-end framework; 2) the design of a location network, which is based on reinforcement learning and aims to learn task-specific sequential spatial correspondences for different image pairs through the local pairwise internal representation interactions. The proposed model is evaluated on three benchmarks, including Market-1501, DukeMTMC-reID and CUHK03, and achieves better performances than other methods.

Electrical capacitance tomography (ECT) has been widely used in fluidized bed measurements because of the noninvasive and nonintrusive features. The successful application of ECT in fluidized bed measurements depends on the appropriate design of sensors. This paper discusses the effect of ECT sensor length on fluidized bed measurements by simulations and experiments. The sensitivity distributions in different axial positions of a sensor are compared, which shows the similarity of sensitivity distribution contours and decline of intensity along the sensor axial direction. Fluidized beds of three different diameters (45, 55, and 65 mm) are used to compare the effect of sensor electrode length on fluidized bed measurements. In each fluidized bed, five sensors with different electrode lengths are tested. The experimental results show that the sensor electrode length has significant effect on bubble detection in fluidized bed measurements, and it is recommended that the sensor electrode length should be as short as possible considering the signal-to-noise ratio constraint.

For the original generative adversarial networks (GANs) model, there are three problems that (1) the generator is not robust to the input random noise; (2) the discriminating ability of discriminator gradually reduces in the later stage of training; and (3) it is difficult to reach at the theoretical Nash equilibrium point in the process of training. To solve the above problems, in this paper, a GANs model with denoising penalty and sample augmentation is proposed. In this model, a denoising constraint is firstly designed as the penalty term of the generator, which minimizes the F-norm between the input noise and the encoding of the image generated by the corresponding perturbed noise, respectively. The generator is forced to learn more robust invariant characteristics. Secondly, we put forward a sample augmentation discriminator to improve the ability of discriminator, which is trained by mixing the generated and real images as training samples. Thirdly, in order to achieve the theoretical optimization as far as possible, our model combines denoising penalty and sample augmentation discriminator. Then, denoising penalty and sample augmentation discriminator are applied to five different GANs models whose loss functions include the original GANs, Hinge and least squares loss. Finally, experimental results on the LSUN and CelebA datasets show that our proposed method can help the baseline models improve the quality of generated images.

Determination of the amino acid phenylalanine is important for lifelong disease management in patients with phenylketonuria, a genetic disorder in which phenylalanine accumulates and persists at levels that alter brain development and cause permanent neurological damage and cognitive dysfunction. Recent approaches for treating phenylketonuria focus on injectable medications that efficiently break down phenylalanine but sometimes result in detrimentally low phenylalanine levels. We have identified new DNA aptamers for phenylalanine in two formats, initially as fluorescent sensors and then, incorporated with field-effect transistors (FETs). Aptamer-FET sensors detected phenylalanine over a wide range of concentrations (fM-mM). para-Chlorophenylalanine, which inhibits the enzyme that converts phenylalanine to tyrosine, was used to induce hyperphenylalaninemia during brain development in mice. Aptamer-FET sensors were specific for phenylalanine vs para-chlorophenylalanine and differentiated changes in mouse serum phenylalanine at levels expected in patients. Aptamer-FETs can be used to investigate models of hyperphenylalanemia in the presence of structurally related enzyme inhibitors, as well as naturally occurring amino acids. Nucleic acid-based receptors that discriminate phenylalanine analogs, some that differ by a single substituent, indicate a refined ability to identify aptamers with binding pockets tailored for high affinity and specificity. Aptamers of this type integrated into field-effect transistors enable rapid, electronic, label-free phenylalanine sensing.

In the original GANs model, there exist two drawbacks for image generation task. On the one hand, a generator cannot deal with a noise that causes an abrupt loss change and inputs of these noises also affect a discriminator, resulting in instability in training process. On the other hand, the discriminator’s discriminating ability is limited in the later stage of training. Eventually, generated images are blurred and targets are incomplete. In order to solve the above problems, a GANs model with hybrid augmented discriminator and fake sample penalty is firstly proposed. In this model, we design a hybrid augmented discriminator. We add real and fake samples into this discriminator. These hybrid samples are conducive to improve the discriminating ability of discriminator. Then, to stabilize the training process and achieve local convergence, we add a penalty to the generator on the basis of designed discriminator, which constrains fake samples with ill condition number. Secondly, we validate the generalization of proposed method on three different loss functions including Hinge, GANs and LSGANs loss. Finally, experimental results show that the proposed method is more effective than baseline models.

In view of the significance of non-spherical and permeable particles in liquid-solid and gas-liquid-solid reactors in industrial processes, it is essential to understand and quantify the rheological properties of multiphase flows in these processes. In this study, we investigate the shear viscosity of dilute suspension containing elliptical porous particles at low Reynolds number Re of O(0) by use of a modified lattice Boltzmann model. The fluid flow around and inside an elliptical porous particle is described by the volume-averaged macroscopic governing equations. The relative viscosity is calculated for an elliptical porous particle rotating in a two-dimensional (2D) simple shear flow, based on the relation between the shear stress and the second order moments of non-equilibrium particle distribution function. The effects of porous structure of the elliptical particle on the viscosity and flow field are investigated with different axis ratios in detail. Our results demonstrate that the relative viscosities of dilute suspension containing elliptical porous particles increase linearly with solid volume fraction at various Darcy number for particles with varying axis ratios. Moreover, a simple empirical expression for intrinsic viscosity is proposed as a function of Darcy number.

A deep convolutional neural network (CNN) becomes a widely used tool for object detection. Many previous works have achieved excellent performance on object detection benchmarks. However, these works present generic detectors whose performance will drop rapidly when they are applied to a surveillance scene. In this paper, we propose an efficient method to construct a scene-specific regression model based on a generic CNN-based classifier. Our regression model is an adaptive deep CNN (ADCNN), which can predict object locations in the surveillance scene. First, we transfer the generic CNN-based classifier to the surveillance scene by selecting useful kernels. Second, we learn the context information of the surveillance scene in our regression model for accurate location prediction. Our main contributions are: 1) a transfer learning method that selects useful kernels in the generic CNN-based classifier; 2) a special architecture that can effectively learn the local and global context information in the surveillance scene; and 3) a new objective function to effectively train parameters in ADCNN. Compared with some state-of-the-art models, ADCNN achieves the best performance on three surveillance data sets for pedestrian detection and one surveillance data set for vehicle detection.

Oxygen nonstoichiometry plays a critical role in determining the physical and chemical functionalities of oxide materials. For widespread applications involving oxygen transport and exchange with t...

China’s unique energy reserve structure abundant in coal and scarce in crude oil and natural gas has promoted heavy investment on the research and development of clean coal chemical technologies during last two decades, which has turned China into a heartland for demonstrating, developing, and commercializing virtually every aspect of new coal chemical process technologies. Consequently, breakthroughs in coal gasification, indirect and direct coal-to-liquid (CTL) processes, and methanol-to-olefins (MTO) technologies are catching attention worldwide. Gasification technology for syngas production is the key to high plant availability and economic success for most coal chemical projects. During the past 20 years, both international and Chinese gasifier vendors have reaped great successes in licensing their technologies in the domestic market. Notably, the local vendors have been investing heavily on inventing and improving their technologies to suit the specific requirement of gasifying a variety of coals. The opposed multinozzle gasification technology from East China University of Science and Technology was taken as an example to demonstrate the recent development in this field. The coal chemical industry in China has witnessed several notable achievements in chemical engineering progress, namely CTL (indirect and direct) and MTO. Comprehensive reviews on topics such as catalysis, kinetics, and reactor design and process integration will be provided by leading scientists in related fields with firsthand information to showcase the contributions of Chinese researchers to chemical engineering science and technology.

This paper presents an experimental investigation for thermal properties of a specialized electric heating concrete containing steel fiber and graphite (SGEHC) which is a specialized concrete used for outdoor snow melting and indoor radiant heating. Experiments were carried out to determine the density, special heat capacity and thermal conductivity of the SGEHC. It shows that the SGEHC is a porous material with 11.7% porosity ratio, which has a great influence on the density, thermal conductivity of the SGEHC but limited effect on its specific heat capacity. Accordingly, a density model and a thermal conductivity model were proposed by treating the SGEHC as a four phases composite made of steel fiber, graphite, regular concrete and air, which could conducts heat in both parallel and series. It shows that the predicted results are in good agreement with the experimental results.

Heterogeneous longitudinal data have become prevalent in medical, biological, and social studies. This paper proposes a finite mixture of varying coefficient models for handling heterogeneous populations. Each component of the mixture is modeled by a varying coefficient mixed-effect model that characterizes the longitudinal relations among variables. The identifiability of the mixture model is studied. Regression splines with equally spaced knots are used to approximate the varying coefficient functions, and a nested expectation maximization algorithm is developed to obtain the maximum likelihood estimation. We propose a penalized likelihood method based on the smoothly clipped absolute deviation (SCAD) penalty for the component selection of finite mixture of varying coefficient model. A modified BIC-based criterion based on the SCAD penalty, the BIC SCAD , is proposed for selecting the penalty parameter and spline space simultaneously. The asymptotic properties of parameter estimation and component selection consistency are studied under mild conditions. Simulation studies are conducted to illustrate the component selection, parameter estimation, and inference of the proposed method. The model is then applied to a heterogeneous longitudinal data set from a study of the treatment effect on the use of heroin in the California Civil Addict Program.

In this paper, we successfully constructed a ternary metal/non-metal nanomaterial which can synergize Z-scheme heterojunction and plasmonic metal/semiconductor for enhancing visible light photoreduction of aqueous Cr(VI). The as-formed ternary metal/non-metal nanomaterial was composed of g-C3N4 nanosheets mutual modified by Ag nanoparticles and Bi4O7. In the ternary nanomaterial, g-C3N4 nanosheets and Bi4O7 fabricated Z-scheme heterojunction, g-C3N4 nanosheets and Ag nanoparticles formed plasmonic metal/semiconductor system, synchronously. The photodeposition experiments were designed to indirect demonstrate the Z-scheme heterojunction via photocatalytic oxidation and reduction deposition reactions, AgNO3 was used as the electron acceptor and Pb(NO3)2 as the electron donor, photocatalytic reduction product and oxidation product were selectively deposited on g-C3N4 nanosheets and Bi4O7, respectively, revealing the Z-scheme heterojunction was definite formed. In addition, X-ray photoelectron spectroscopy (XPS), photoluminescence (PL), UV–vis absorption, and Fourier-transform infrared spectroscopy (FTIR) were studied to indicate that the plasmonic metal/semiconductor system was existent in the ternary nanomaterial. The visible light photoreduction of ternary Ag/Bi4O7/g-C3N4 nanosheets was evaluated by photoreducing aqueous Cr(VI). The photoreduction results show that the ternary Ag/Bi4O7/g-C3N4 nanosheets is much higher than the pristine bulk g-C3N4 and g-C3N4 nanosheets.

HfO2-based unconventional ferroelectric materials were recently discovered and have attracted a great deal of attention in both academia and industry. The growth of epitaxial Si-doped HfO2 films ha...

Due to the small scale, low technical level, and limited scope of application of artificial intelligence in the field of traditional Chinese medicine orthopedics, there are many problems in the development of this area. Driven by precise and minimally invasive medical concepts, the integration of artificial intelligence and data mining technologies (such as the Internet, cloud computing, and big data) will further profoundly change the face of smart healthcare in the future, which will bring great challenges to traditional Chinese medicine orthopedics, and Change opportunities for your own development. In order for artificial intelligence to provide good diagnosis and treatment services for a large number of patients with orthopaedic diseases, progress still needs to be made with the technology and joint efforts of many orthopaedic experts and scholars. This article uses data mining techniques to optimize orthopedic rehabilitation information. Test results show that the system is faster and more convenient. Artificial intelligence technology will be used for intelligent medical treatment of orthopedic diseases in the future. It will become an essential assistant for clinicians and make a significant contribution to the diagnosis of various orthopedic diseases.

Traditional object detection methods always assume both of the training and test data follow the same distribution, but this cannot always be guaranteed in the real world. Domain adaptive methods are proposed to handle this situation. However, existing methods generally ignore the semantic alignment at feature level when they try to align data distributions between source and target domains. In this paper, we propose a novel unsupervised cross-domain object detection method, named Cycle-consistent domain Adaptive Faster RCNN (CA-FRCNN). A couple of Generative Adversarial Nets (GAN) are used to make the features from two domains consistent at both data distribution level and semantic level. Specifically, features from source domain are transformed to the target domain. Then they are aligned with features from target domain. At the same time, target features are handled with similar operations. Furthermore, a cycle-consistent loss is optimized to guarantee that the semantic information is preserved before and after the style translations. In the end, identity module is used to make the feature in source domain equivalent to the reconstructed feature output by the source generator, whose input is from the source domain. For the feature in the target domain, similar identity is required. Experiments on multiple datasets show that our method performs better than previous state-of-the-art methods.

Person re-identification (re-id) refers to matching people across disjoint camera views. Most of person re-id methods extract discriminative features from the whole images or fixed regions and develop their metrics. However, these methods ignore that the attention regions with temporal cues in the pedestrian image pair hold discriminative information. In this paper, we propose the recurrent models of visual co-attention that aim to simulate human eye movement, focusing on the sequential concurrent attention (co-attention) regions of the same locations when comparing image pairs. Since reinforcement learning provides a flexible learning strategy for sequential decision-making, it is naturally applied to perform the temporal re-id co-attention learning task. The reward functions are designed to recursively optimize the prediction by rewarding or punishing the learning process. The recurrent models are used to extract information from a sequence of attention regions. Finally, person re-id is performed based on the whole image feature and the features from the recurrent models. Our contributions are: 1) the visual mechanism, which can dynamically locate the optimal co-attention regions to simulate the human re-id process; 2) the design of reward functions in reinforcement learning, which aims to recursively optimize the prediction process; and 3) experimental results, which demonstrate the advantages of our method compared with the state-of-the-art methods.

Most existing human pose estimation methods focus on enhancing the accuracy performance alone while ignoring the critical model efficiency issue. This dramatically limits their scalability and deployability in large-scale applications. In this paper, we consider the under-studied model efficiency problem in pose estimation. We demonstrate the advantages and potential of hierarchical context learning in the convolutional neural network. Specifically, we formulate a novel hierarchical context network (HCN) architecture that can be trained and deployed efficiently while achieving competitive model generalization capability. This is achieved by progressively forming and imposing multi-granularity context information during the pose regression learning process in a coarse-to-fine manner. The extensive comparative evaluations validate the superiority of the proposed HCN over a wide variety of the state-of-the-art human pose estimation models on two challenging benchmarks: MPII and LSP.

Organochlorine pesticides (OCPs) contaminated sites pose great threats to both human health and environmental safety. Targeted bioremediation in these regions largely depends on microbial diversity and activity. This study applied metagenomics to characterize the microbial communities and functional groups composition features during independent or simultaneous rapeseed oil and tartaric acid applications, as well as the degradation kinetics of OCPs. Results showed that: the degradation rates of α-chlordane, β-chlordane and mirex were better when (0.50% w/w) rapeseed oil and (0.05 mol L-1) tartaric acid were applied simultaneously than singular use, yielding removal rates of 56.4%, 53.9%, and 49.4%, respectively. Meanwhile, bio-stimulation facilitated microbial enzyme (catalase/superoxide dismutase/peroxidase) activity in soils significantly, promoting the growth of dominant bacterial communities. Classification at phylum level showed that the relative abundance of Proteobacteria was significantly increased (p < 0.05). Network analysis showed that bio-stimulation substantially increased the dominant bacterial community's proportion, especially Proteobacteria. The functional gene results illustrated that bio-stimulation facilitated total relative abundance of degradation genes, phosphorus, carbon, nitrogen, sulfur metabolic genes, and iron transporting genes (p < 0.05). In metabolic pathways, functional genes related to methanogenesis and ammonia generation were markedly upregulated, indicating that bio-stimulation promoted the transformation of metabolic genes, such as carbon and nitrogen. This research is conducive to exploring the microbiological response mechanisms of bio-stimulation in indigenous flora, which may provide technical support for assessing the microbial ecological remediation outcomes of bio-stimulation in OCP contaminated sites.

Although epitaxial strain imparted by lattice mismatch between a film and the underlying substrate has led to distinct structures and emergent functionalities, the discrete lattice parameters of limited substrates, combined with strain relaxations driven by film thickness, result in severe obstructions to subtly regulate electro‐elastic coupling properties in perovskite ferroelectric films. Here a practical and universal method to achieve highly strained phases with large tetragonal distortions in Pb‐based ferroelectric films through synergetic effects of moderately (≈1.0%) misfit strains and laser fluences during pulsed laser deposition process is demonstrated. The phase possesses unexpectedly large Poisson's ratio and negative thermal expansion, and concomitant enhancements of spontaneous polarization (≈100 µC cm−2) and Curie temperature (≈800 °C), 40% and 75% larger than that of bulk counterparts, respectively. This strategy efficiently circumvents the long‐standing issue of limited numbers of discrete substrates and enables continuous regulations of exploitable lattice states in functional oxide films with tightly elastic coupled performances beyond their present levels., A highly distorted tetragonal phase with giant negative thermal expansions is achieved in Pb‐based ferroelectric films via synergistic effects of misfit strainand laser fluence, which possesses considerable enhancements of spontaneous polarization (≈100 µC cm−2) and Curie temperature (≈800 °C). This strategy efficiently circumvents the issue of limited substrates and enables continuous regulations of ferroelectric films with tightly elastic coupled functionality.

A conventional electrical capacitance tomography (ECT) sensor consists of multiple rectangular electrodes in a single plane. To generate 3-D images, some researchers use an ECT sensor with multiplane rectangular electrodes and in-line arrangement of different electrode planes and measure the capacitance between the electrode pairs not only within the same plane but also between different planes. Because those electrode pairs in different planes are far away from each other, the measured capacitance signals between them are weak and, hence, are easily affected by noise, resulting in the obvious disturbance on the measured signals and, hence, low image quality. To address those problems, the 3-D ECT sensors with different arrangements and different shapes of electrodes are studied to investigate the effect of different arrangements and different shapes of electrodes on the measured capacitance signals and the image quality. Four 3-D ECT sensors are made with different arrangements or shapes of electrodes: rectangular electrodes are in line arrangement, whereas diamond, circular, and hexagon electrodes are staggered arrangement. The performance of the four 3-D ECT sensors is compared. The results show that 3-D ECT sensors with staggered electrodes can decrease the disturbance on the measured signals and improve the image quality of static models compared with that with conventional rectangular electrodes. Among the three 3-D ECT sensors with staggered electrodes, the 3-D ECT sensor with hexagonal electrodes has superior performance to the other two 3-D ECT sensors with diamond and circular electrodes. It was used to measure moving objects and a fluidized bed.

The current encoding options not only decide the rate and distortion of the current coding unit or frame, but also may affect the achievable rate-distortion (R-D) performance for the subsequent coding units or frames. In order to measure quantitatively the temporal R-D dependency, a distortion effect factor is first introduced, which can be calculated with the coding distortion and motion compensation predicted (MCP) error. Then an adaptive Lagrange multiplier determination at coding tree unit (CTU) level is proposed based on a two-pass encoding framework. The framework mainly includes three parts. First, pre-encoding a frame with a simplified RDO to obtain the distortion effect factors. Second, restoring the reference list to the status before pre-encoding. Third, calculating the Lagrange multipliers for each CTU according to the distortion effect factors obtained, and encoding the previous frame again. Experimental results show that, on the platform of HEVC test model (HM-16.7), the proposed algorithm obtains bitrate savings of 1.8% and 1.7% on average under the configurations of low delay B and P frames, with increased encoding complexity of 23% on average.

Absence of nearby light sources while capturing an image will degrade the visibility and quality of the captured image, making computer vision tasks difficult. In this paper, a color-wise attention network (CWAN) is proposed for low-light image enhancement based on convolutional neural networks. Motivated by the human visual system when looking at dark images, CWAN learns an end-to-end mapping between low-light and enhanced images while searching for any useful color cues in the low-light image to aid in the color enhancement process. Once these regions are identified, CWAN attention will be mainly focused to synthesize these local regions, as well as the global image. Both quantitative and qualitative experiments on challenging datasets demonstrate the advantages of our method in comparison with state-of-the-art methods., Comment: 8 pages, 9 figures

While being the de facto standard coordinate representation in human pose estimation, heatmap is never systematically investigated in the literature, to our best knowledge. This work fills this gap by studying the coordinate representation with a particular focus on the heatmap. Interestingly, we found that the process of decoding the predicted heatmaps into the final joint coordinates in the original image space is surprisingly significant for human pose estimation performance, which nevertheless was not recognised before. In light of the discovered importance, we further probe the design limitations of the standard coordinate decoding method widely used by existing methods, and propose a more principled distribution-aware decoding method. Meanwhile, we improve the standard coordinate encoding process (i.e. transforming ground-truth coordinates to heatmaps) by generating accurate heatmap distributions for unbiased model training. Taking the two together, we formulate a novel Distribution-Aware coordinate Representation of Keypoint (DARK) method. Serving as a model-agnostic plug-in, DARK significantly improves the performance of a variety of state-of-the-art human pose estimation models. Extensive experiments show that DARK yields the best results on two common benchmarks, MPII and COCO, consistently validating the usefulness and effectiveness of our novel coordinate representation idea., Results on the COCO keypoint detection challenge: 78.9% AP on the test-dev set (Top-1 in the leaderbord by 12 Oct 2019) and 76.4% AP on the test-challenge set. Project page: https://ilovepose.github.io/coco

Pyrite-structured PdSb2 with the nonsymmorphic crystalline symmetry, has long been predicted to host sixfold-degenerate exotic fermions beyond Dirac and Weyl. Although magnetotransport measurements on PdSb2 have suggested its topologically nontrivial character, the direct spectroscopic evidence still remains absent. By utilizing high-resolution angle-resolved photoemission spectroscopy, we present a systematic study on the low-energy bulk and surface electronic structure of pyrite-structured PdSb2. Through careful comparison with first-principles calculations, we verify the existence of sixfold fermions in this compound, which are formed by three doubly degenerate bands centered at the R point in the Brillouin zone. These bands exhibit parabolic dispersion close to sixfold fermion nodes, in sharp contrast to previously reported sixfold double spin-1 chiral fermions. Furthermore, our findings reveal no protected Fermi arcs in PdSb2, which is compatible with its achiral structure. Our findings suggest that pyrite-structured PdSb2 provides an ideal platform for the investigation of fermions and their potential applications.

This study quantifies the impact of individual attributes, the built environment, and travel characteristics on the use of bike-sharing and the willingness of shifting to bike-sharing-related travel modes (bike-sharing combined with other public transportation modes such as bus and subway) under different scenarios. The data are from an RP (Revealed Preference) survey and SP (Stated Preference) survey in Nanjing, China. Three mixed logit models are established: an individual attribute&ndash, travel characteristics model, a various-factor bike-sharing usage frequency model, and a mixed scenario&ndash, transfer willingness model. It is found that age and income are negatively associated with bike-sharing usage, the transfer distance (about 1 km), owning no car, students, and enterprises are positively associated with bike-sharing usage, both weather and travel distance have a significant negative impact on mode shifting. The sesearch conclusions can provide a reference for the formulation of urban transportation policies, the daily operation scheduling, and service optimization of bike-sharing.

Chalcogenide phase-change materials show strikingly contrasting optical and electrical properties, which has led to their extensive implementation in various memory devices. By performing spin-, time-, and angle-resolved photoemission spectroscopy combined with the first-principles calculation, we report the experimental results that the crystalline phase of GeSb2Te4 is topologically nontrivial in the vicinity of the Dirac semimetal phase. The resulting linearly dispersive bulk Dirac-like bands that cross the Fermi level and are thus responsible for conductivity in the stable crystalline phase of GeSb2Te4 can be viewed as a 3D analogue of graphene. Our finding provides us with the possibility of realizing inertia-free Dirac currents in phase-change materials., The ARPES experiments were performed with the approval of the Proposal Assessing Committee of HSRC (Proposal No. 18BG039). The TARPES measurements were jointly carried out by the Laser and Synchrotron Research Center of the Institute for Solid State Physics at the University of Tokyo. This work was financially supported by KAKENHI (Grant No. 17H06138, No. 17H06152, No. 18H03683, No. 18H01148). This work was also supported by the ‘‘Tomsk State University competitiveness improvement programme’’ (Project No. 8.1.01.2018), by Saint Petersburg State University (Project ID 51126254), and by Russian Science Foundation No. 17-12-01047 (in the crystal growth part). T.V.M. and I.P.R. acknowledge support from Ministry of Education and Science of the Russian Federation (State Task No. 0721-2020-0033) (study of intermixed crystalline phase of GST-124). Calculations were performed at the SKIF-Cyberia supercomputer of Tomsk State University (Russian Federation). K.A.K. acknowledges RFBR Grant 17-08-00955 and a state contract of IGM SB RAS.

Generative adversarial networks (GANs) have attracted much attention since it is able to effective learn from an unknown real distribution. However, the instability of the training process greatly affects the quality of the generated images. To address this problem, the network structure-based, loss-based variant model and some training techniques are proposed. Unfortunately, there are some problems with the above methods, such as the limited effect of stabilizing the training process, the complex mathematical derivation, and the lack of universality of training techniques for different tasks. To this end, we propose a novel squeeze criterion GANs. In this method, we design a pseudo real module to synthesize adversarial sample and the double identity discriminator is designed. Then, the generated image and adversarial sample, as well as the generated image and real image form double adversarial learning. Through double adversarial learning, it forms a squeeze criterion to stabilize the training process of generator and discriminator. Finally, experimental results show that the proposed method has well portability and stabilizes the training process of existing GANs, and improves the quality of generated images.

State-of-the-art NLP models can often be fooled by human-unaware transformations such as synonymous word substitution. For security reasons, it is of critical importance to develop models with certified robustness that can provably guarantee that the prediction is can not be altered by any possible synonymous word substitution. In this work, we propose a certified robust method based on a new randomized smoothing technique, which constructs a stochastic ensemble by applying random word substitutions on the input sentences, and leverage the statistical properties of the ensemble to provably certify the robustness. Our method is simple and structure-free in that it only requires the black-box queries of the model outputs, and hence can be applied to any pre-trained models (such as BERT) and any types of models (world-level or subword-level). Our method significantly outperforms recent state-of-the-art methods for certified robustness on both IMDB and Amazon text classification tasks. To the best of our knowledge, we are the first work to achieve certified robustness on large systems such as BERT with practically meaningful certified accuracy., Comment: ACL 2020

The methanol-to-olefins (MTO) reaction over the SAPO-34 catalyst suffers rapid deactivation due to coke deposition. Here, partial regeneration methods with air combustion and steam gasification were investigated. The results showed that both methods could not only reactivate the spent SAPO-34 catalyst but also prompt ethylene selectivity. Whereas, steam gasification was more effective , by which the initial ethylene selectivity could reach 53%, much higher than that of 23% over fresh SAPO-34 catalyst. Several characterizations, e.g., Fourier transform infrared spectroscopy and gas chromatography–mass spectrometry, indicated that methylated benzenes, which were active hydrocarbon pool species and prompted ethylene selectivity, presented as the main species of residual coke in the SAPO-34 catalyst regenerated via steam gasification. Meanwhile, oxygenated compounds and naphthalene, which could accelerate catalyst deactivation, were observed in the SAPO-34 catalyst partially regenerated via air combustion. Th...