Search

Cross-domain images have been witnessed in an increasing number of applications. This new trend triggers demands for cross-domain image retrieval (CDIR), which finds images in one visual domain according to a query image from another visual domain. Although image retrieval has been studied extensively, exploration of the CDIR remains at its initial stage. This study systematically surveys the methods and applications of the CDIR. Since images from different visual domains exhibit different features, learning discriminative feature representations while preserving domain-invariant features of images from different visual domains is the main challenge of the CDIR. According to the feature transformation stage of images from different visual domains, existing CDIR methods are categorized and analyzed. One is based on feature space migration and the other is based on image domain migration. Then, applications of CDIR in clothing, infrared, remote sensing, sketch, and other scenarios are summarized. Finally, the existing CDIR schemes are concluded, and new directions for future research are proposed.

Graph anomaly detection (GAD) is an emerging and essential research field for discovering anomalous individuals (e.g., nodes or edges) that deviate significantly from the normal majority in an attributed graph. Unlike other anomaly types (e.g., image, text) with independence, it is not trivial to capture inherent and distinctive anomaly patterns as graph anomalies usually exist more complex relational interaction with a overwhelming amount of normal majority, which induces the inconsistency in feature and structure space. Recently, some studies have applied graph self-supervised learning on the GAD task and surge a new climax. Despite their success, the sub-optimal performance is only achieved because the imbalance nature of anomaly problem significantly dilutes anomalous information and causes the shortage of effective supervision signals. To address these challenges, we propose a dual-bootstrapped self-supervised approach, namely DualGAD, that consists of one generative module and one contrastive module. Specifically, we first sample a local subgraph for each target node and employ two mask strategies on subgraphs to break the short-range connection and reduce redundancy spread from majority neighbors. The generative module is equipped with the reconstruction objective to model the surrounding context of masked subgraphs, which learn discriminative representations by capturing the inconsistency patterns at both feature and structure space. Considering the local information is easily been over-emphasized, we elaborately tailor a novel cluster-guided contrastive learning module without relying on positive/negative pairs to learn the intrinsic anomaly property by aligning the perturbation distributions across views. Finally, the two self-supervised modules are seamlessly integrated and bootstrap each other for learning the discriminative representations to distinguish anomalies and normal majority. Extensive experiments demonstrate the superiority of DualGAD on the node-level anomaly detection task. [ABSTRACT FROM AUTHOR]

ABSTRACTA cross-sectional field survey was conducted from November 2020 to January 2021 among 7259 participants to investigate the public perception, willingness, and information sources for COVID-19 vaccination, with the focus on the elderly and non-communicable chronic disease (NCD) population. Multiple logistic regressions were performed to identify associated factors of the vaccination willingness. The willingness rate of the elderly to accept the future COVID-19 vaccine (79.08%) was lower than that of the adults aged 18–59 (84.75%). The multiple analysis didn't identify significant relationship between NCD status and the vaccination intention. The main reasons for vaccine hesitancy by the public were: concern for vaccine safety, low infection risk, waiting and seeing others getting vaccinated, concern of vaccine effectiveness and price. Their relative importance differed between adults aged 18–59 and the elderly, and between adults aged 18–59 with or without NCD. Perception for vaccination importance, vaccine confidence, and trust in health workers were significant predictors of the vaccination intention in both age groups. The elderly who perceived high infection risk or had trust in governments were more likely to accept the vaccine. Compared with the adults aged 18–59, the elderly used fewer sources for COVID-19 vaccination information and more trusted in traditional media and family, relatives, and friends for getting vaccination recommendations. To promote vaccine uptake, the vaccination campaigns require comprehensive interventions to improve vaccination attitude, vaccine accessibility and affordability, and tailor strategies to address specific concerns among different population groups and conducted via their trusted sources, especially for the elderly.

A least squares support vector machine (LS-SVM) offers performance comparable to that of SVMs for classification and regression. The main limitation of LS-SVM is that it lacks sparsity compared with SVMs, making LS-SVM unsuitable for handling large-scale data due to computation and memory costs. To obtain sparse LS-SVM, several pruning methods based on an iterative strategy were recently proposed but did not consider the quantity constraint on the number of reserved support vectors, as widely used in real-life applications. In this article, a noniterative algorithm is proposed based on the selection of globally representative points (global-representation-based sparse least squares support vector machine, GRS-LSSVM) to improve the performance of sparse LS-SVM. For the first time, we present a model of sparse LS-SVM with a quantity constraint. In solving the optimal solution of the model, we find that using globally representative points to construct the reserved support vector set produces a better solution than other methods. We design an indicator based on point density and point dispersion to evaluate the global representation of points in feature space. Using the indicator, the top globally representative points are selected in one step from all points to construct the reserved support vector set of sparse LS-SVM. After obtaining the set, the decision hyperplane of sparse LS-SVM is directly computed using an algebraic formula. This algorithm only consumes O(N2) in computational complexity and O(N) in memory cost which makes it suitable for large-scale data sets. The experimental results show that the proposed algorithm has higher sparsity, greater stability, and lower computational complexity than the traditional iterative algorithms.

The semisupervised least squares support vector machine (LS-S3VM) is an important enhancement of least squares support vector machines in semisupervised learning. Given that most data collected from the real world are without labels, semisupervised approaches are more applicable than standard supervised approaches. Although a few training methods for LS-S3VM exist, the problem of deriving the optimal decision hyperplane efficiently and effectually has not been solved. In this paper, a fully weighted model of LS-S3VM is proposed, and a simple integer programming (IP) model is introduced through an equivalent transformation to solve the model. Based on the distances between the unlabeled data and the decision hyperplane, a new indicator is designed to represent the possibility that the label of an unlabeled datum should be reversed in each iteration during training. Using the indicator, we construct an extended candidate set consisting of the indices of unlabeled data with high possibilities, which integrates more information from unlabeled data. Our algorithm is degenerated into a special scenario of the previous algorithm when the extended candidate set is reduced into a set with only one element. Two strategies are utilized to determine the descent directions based on the extended candidate set. Furthermore, we developed a novel method for locating a good starting point based on the properties of the equivalent IP model. Combined with the extended candidate set and the carefully computed starting point, a fast algorithm to solve LS-S3VM quasi-optimally is proposed. The choice of quasi-optimal solutions results in low computational cost and avoidance of overfitting. Experiments show that our algorithm equipped with the two designed strategies is more effective than other algorithms in at least one of the following three aspects: 1) computational complexity; 2) generalization ability; and 3) flexibility. However, our algorithm and other algorithms have similar levels of performance in the remaining aspects. [ABSTRACT FROM AUTHOR]

The semisupervised least squares support vector machine (LS-S3VM) is an important enhancement of least squares support vector machines in semisupervised learning. Given that most data collected from the real world are without labels, semisupervised approaches are more applicable than standard supervised approaches. Although a few training methods for LS-S3VM exist, the problem of deriving the optimal decision hyperplane efficiently and effectually has not been solved. In this paper, a fully weighted model of LS-S3VM is proposed, and a simple integer programming (IP) model is introduced through an equivalent transformation to solve the model. Based on the distances between the unlabeled data and the decision hyperplane, a new indicator is designed to represent the possibility that the label of an unlabeled datum should be reversed in each iteration during training. Using the indicator, we construct an extended candidate set consisting of the indices of unlabeled data with high possibilities, which integrates more information from unlabeled data. Our algorithm is degenerated into a special scenario of the previous algorithm when the extended candidate set is reduced into a set with only one element. Two strategies are utilized to determine the descent directions based on the extended candidate set. Furthermore, we developed a novel method for locating a good starting point based on the properties of the equivalent IP model. Combined with the extended candidate set and the carefully computed starting point, a fast algorithm to solve LS-S3VM quasi-optimally is proposed. The choice of quasi-optimal solutions results in low computational cost and avoidance of overfitting. Experiments show that our algorithm equipped with the two designed strategies is more effective than other algorithms in at least one of the following three aspects: 1) computational complexity; 2) generalization ability; and 3) flexibility. However, our algorithm and other algorithms have similar levels of performance in the remaining aspects.

Ancient Chinese cartography dates back to the Western Jin Dynasty in the third century. Cartography was initiated during this period by Pei Xiu, a minister and cartographic expert, who suggested six principles of cartography. Pei Xiu, who is known as the father of Chinese cartography, oversaw the completion of the “Yu Gong Regional Maps”, along with 18 articles and the “Terrain Fangzhang Map” (AD 224–271). This led to a number of subsequent cartographic initiatives including the “Wooden Fangzhang Map” (an administrative map) which was completed by Xie Zhuang, a minister during the Southern Dynasties (AD 502–557), “Haineihuayi Tu” (Map of China and its neighbouring countries) (AD 730–805), drawn by Jia Dan, a cartographer of the Tang Dynasty, and “Shouling Tu” (an administrative map, AD 1031–1095), drawn by Shen Kuo, a scientist during the Northern Song Dynasty. Throughout the 16th century, ancient Chinese cartography developed continuously and cumulatively, and formed the specific characteristics of China ancient cartography. Although Western latitude- based and longitude-based mapping techniques introduced to China in AD 1460, the theories and methods of Chinese and Western mapping systems co-existed for over 400 years. The cultural heritage of Chinese cartography can be seen in many famous ancient Chinese maps, including astronomical figures and atlases (world, Chinese, regional, military, water conservancy, historical, and educational maps), charts, and maps of scenic spots. These have hitherto been kept in well-known archives and institutions across the world. They form an important part of the global cultural heritage of ancient maps and cartography. Given their high cultural value, these maps remain an important point of study. This paper provides a preliminary discussion on the rarity, application, and the historical, scientific, and artistic value of ancient Chinese maps.

In this paper, knock detection and analysis methods of the gasoline engine are applied to a modified kerosene engine. Two kinds of signals, the cylinder pressure signal and the vibration signal, are acquired by a pressure sensor mounted in a spark plug and an accelerometer mounted on the cylinder block separately. Due to the influence of noise, such as cam shocks, piston up-and-down motion, crank rotation and so on, the vibration signal should be processed to extract knock feature. Firstly, FFT, a frequency analysis approach, is used to determine the characteristic frequency range of knock. Then, combining the sampling frequency, the vibration signal is divided into different frequency sub-bands, and the characteristic sub-band of DWT, a time-frequency analysis method, is determined. After that, DWT is applied to extract knock feature from the vibration signal. At last, the knock energy evaluation index is used to judge knock intensity. The reference knock energy is chosen by the researchers, and it is not absolutely suitable for the engine operating condition, so intuitive cylinder pressure signal is applied to supplement for knock judgment.

Mine environmental geological assessment usually uses the expert subjective assignment to determine the weights of affecting factors, but the mine environmental geological problems are the result of many factors. Each factor is complex on the suitability of the influence factors, and the factors interrelate and restrict each other, so the human factor is of great impact. Take the geological quality evaluation of the mine environment of Shanxi, Shaanxi and Inner Mongolia mining area as an example, this article discusses in detail the process of determining the weights of evaluation factors by using the analytic hierarchy process (AHP), which increases the scientific nature of the assignment and reduces the subjectivity, which achieves satisfactory results.

We should construct a spatial database system for mineral resource reserve in Hebei province based on the database with the application of GIS spatial database technology based on the MapGIS software platform. Combined with the construction process of the database, this paper summarizes the problems and the solutions encountered in the construction of large database, and analyzes some techniques the departments of land and resources often use in basic geographic data to discuss with colleagues to further improve the quality and efficiency of database construction.

Aiming at the problem that the stability of clustering recognition algorithm without any prior information is not enough and the network cluster extraction quality needs to be improved and so on, the original algorithm is improved from the following aspects: The choice of the destination vertex function is added in the random walk; use automatic step number detection method to cancel the original algorithm of random walk steps parameter; add weight of evaluation inter-cluster connection on the original based on cluster extraction conditions; realize the controllable particle size of cluster extraction by introducing the threshold parameter. The test results show that the improved algorithm in terms of anti-jamming in the stability and the clustering quality are higher than the original algorithm.

Cluster network structure is one of the most common topological properties of complex networks and the most important topological attribute, and has the characteristics of the same cluster nodes connected to each other, and the sparse characteristics of the dense clusters interconnected with nodes. To reveal the complex network clustering method of network cluster structure has a very important theoretical significance for the analysis of complex network topology structure, to understand its function, discovering the behavior of the hidden patterns, prediction, and is widely used in social networks, biological networks and the world wide web. This article reviews the research background of the complex network clustering method, research significance, research status at home and abroad as well as the major problems currently facing this new research direction in an attempt to draw a comprehensive and clear outline, to provide a useful reference for researchers of the complex network, data mining, intelligent Web, Bioinformatics and other related fields.

n order to accurately predict the future trend of the stock index, a stock index prediction model (SIPM) is provided. SIPM composed by the following modules: data acquisition module, data preprocessing module, learning modules and forecasting modules. The core of SIPM is the last module which adopts ANN as its predicting model. Based on the historical data of a past period, SIPM can forecast next M days trends. The experiments on daily, weekly and monthly data show that the model is feasible. The key conclusion is that SIPM has good prediction results on low-level time granularity.

Precision grinding experiment and process optimization of YG3 cemented carbide materials was conducted with various grinding parameters using resin-bonded diamond grinding wheel. Different rotation speed of workpiece, feed rate and grinding depth were chozen in precision cross grinding. Surface roughness was measured for studying the removal mode and grinding charateristics. A set of ideal grinding parameters of YG3 cemented carbide were obtained by changing the process parameters. The research result has guiding significance for reducing the wheel wear and processing costs, improving grinding quality and efficiency while precision grinding hard and brittle materials.

A hierarchical model of survival situational awareness is proposed based on multi-feature fusion, which includes three layers: feature fusion layer, situation assessment layer and comprehensive perception layer. In the feature fusion layer, light-computation features and efficient fusion mechanisms are employed to improve degree of differentiation. Situation assessment layer adopted fusion technology of variable fuzzy set theory to improve the accuracy and objectivity of evaluation. Furthermore, time series of services are introduced to judge the safety states and predict the trends in comprehensive perception layer. Experiments indicate that proposed model is more scientific and more accurate, which is valuable for generalizations and applications.

Relaxation oscillations in Er^3+-doped and Yb^3+/Er^3+ co-doped fiber grating lasers were investigated. Intensity noise models were established for both cases in which the fluctuations of pump power and cavity loss were analyzed. Simulation indicates that the relaxation oscillation is induced by the pump power fluctuation, and fluctuation of the cavity loss can broaden the relaxation oscillation peak and enhance the amplitude. The relaxation oscillation of a laser with Er^3+-doped fiber exhibits a lower frequency and larger amplitude than that of an Yb^3+/Er^3+ co-doped fiber grating laser. Distributed Bragg reflector fiber grating lasers with Er^3+-doped fiber and Yb^3+/Er^3+ co-doped fiber were fabricated, and the relaxation oscillations were measured. For the former, the relaxation oscillation frequency is at the magnitude of several hundreds of kilohertz and lineally shifts toward the higher frequency with the pump power. For the latter, no obvious relaxation oscillation was measured.

Not Available

Micro-orientation data of a high purity Al rolled up to total thickness reduction of 80% at room temperature were determined using SEM-EBSD technique, conceptions of describing substructure information, such as subgrain misorientation (θcry), and average misorientation (θenv) of circumjacent subgrains for a special subgrain, etc., were suggested and corresponding GCDP-OI soft package was developed. It is found that the subgrains sizes increase rapidly from about 2 to 7 μm with increasing misorientations from 1° to 15°, and the total number frequency of which is more than 95%. However, taking into account local features of subgrais, whose sizes for Dcry/Denv > 1 are 2 times larger than that for Dcry/Denv < 1 on the same misorientation levels below 20°, and the relationships between misorientations and sizes are consistent with that if Dcry/Denv > 1, θcry/θenv > 1, vice versa.

Not Available

The mechanical properties and crosslink density of peroxide-cured ethylene-propylene-diene rubber (EPDM) reinforced with zinc dimethacrylate (ZDMA) were studied. ZDMA was in situ prepared in EPDM matrix through the neutralization reaction of zinc oxide (ZnO) and methacrylate acid (MAA). The effect of ZnO/MAA amount and molar ratio of ZnO/MAA on the properties of the EPDM vulcanizate were investigated in detail. The experimental results showed that EPDM can be greatly reinforced by ZDMA. The excess amount of ZnO considerably increases the tensile strength of the EPDM vulcanizate to reach as high as 37 MPa, whereas its elongation at break keeps over 350%. The process of in situ formation of ZDMA in the EPDM compound was verified by WAXD. Such vulcanizate contains both covalent crosslinks and ionic crosslinks. Crosslink density was determined by an equilibrium swelling method. Dependence of crosslink density on the amount and molar ratio of ZnO/MAA was studied and the extraordinary high tensile strength of the EPDM/ZDMA vulcanizate was related to ionic crosslink density. © 2002 Wiley Periodicals, Inc. J Appl Polym Sci 84: 1339–1345, 2002; DOI 10.1002/app.10112

The reinforcement of non-polar ethylene-propylene-diene rubber (EPDM) by the polar metal salts of methacrylic acid was achieved by directly mixing the salt into EPDM at a high level and curing with peroxide. Mechanical properties and crosslink density of the vulcanizates were investigated. Wide-angle X-ray diffraction (WAXD) and Fourier transform-infrared spectroscopy (FTIR) were employed to analyze the reactivity of zinc dimethacrylate (ZDMA) in EPDM compounds before and after vulcanization. It was observed that the commercial ZDMA or magnesium dimethacrylate (MDMA) can greatly improve the hardness and tensile strength of the EPDM vulcanizates, while retaining a high elongation at break. WAXD and FTIR revealed that ZDMA is a reactive filler in the EPDM matrix during peroxide curing. The crosslink density of the vulcanizates was determined by equilibrium swelling. Covalent crosslinks and ionic crosslinks were distinguished by an acidolysis treatment. Ionic crosslinks were greatly improved by increasing the amount of zinc dimethacrylate or dicumyl peroxide, and the relationship between mechanical properties and crosslink density indicates that the tensile strength of the vulcanizate was principally dependent on ionic crosslink density.

Multilayer SiO2 crystals modified by alkoxysilanes were fabricated by the vertical deposition technique combined with self-assembly method. The SiO2 colloidal crystallization was prepared by the traditional Stöber-Fink-Bohn method, and modified by different alkoxysilanes to investigate the properties of the films. The films fabricated by the ordering of SiO2 spheres were investigated by infrared spectrometer, X-ray diffraction, scanning electron microscope, UV-visible spectrophotometer. The results showed that thin film is the face-centered cubic structures and the spheres are [111]-oriented.

TiO2 coatings on ceramic substrate were prepared using TiOSO4 and silica sol. Coatings were then heat-treated respectively at 300°C, 400°C or 500°C for 30mins. Degradation of the solution of chlorinated methane by UV light and TiO2 coating were investigated. It can be concluded from the results that the coating with composition of 80wt%TiO2-20wt%SiO2 heated at 500°C display better degradation effect than the other coatings. As the initial concentration of chloromethane goes up, the degradation rate increases; while as the degree of substitution by chlorine goes up, the catalytic effects decreases.

Sol-gel coatings of mixed TiO2-SiO2 had been deposited on ceramic substrates by spraycoating technique to study their hydrophilic and photocatalytic behavior on exposure to sunlight or ultraviolet illumination. The range of the films heat-treatment temperature is from 300°C to 500°C. AFM and XRD analysis were used to characterize the crystal structure and surface morphology of these coatings. Hydrophilic property was evaluated by water contact angle measurement under sunlight illumination condition, and photocatalytic activity by decomposition rate of aniline blue under ultraviolet illumination. Attrition resistance experiment was also carried out to test the hardness and adherence strength of these films. The experimental results demonstrate that this coating possesses high photo-activity and excellent adhesion to ceramic substrates.

This paper proposes an attention-based LSTM (AT-LSTM) model for financial time series prediction. We divide the prediction process into two stages. For the first stage, we apply an attention model to assign different weights to the input features of the financial time series at each time step. In the second stage, the attention feature is utilized to effectively select the relevant feature sequences as input to the LSTM neural network for the prediction in the next time frame. Our proposed framework not only solves the long-term dependence problem of time series prediction effectively, but also improves the interpretability of the time series prediction methods based on the neural network. In the end of this paper, we conducted experiments on financial time series prediction task with three real-world data sets. The experimental results show that our framework for time series pre-diction is state-of-the-art against the baselines.