In order to learn high-level representation with rich information for dynamic graphs where nodes and edges change dynamically, a dynamic graph convolutional network (DyGCN) is proposed to learn representation as a mixture of both spatial and temporal information. The model performs spatial convolutions to learn structural information on graphs and temporal convolutions to learn historical information along time axis. Besides, the selfadapting mechanism on the spatial convolution layer allows model parameters to update with graphs. Extensive experiments on financial networks for edge classification tasks against financial crimes show that DyGCN outperforms state-of-the-art methods.

To improve the efficiency of lattice-based cryptography in practical applications, the optimization technology of polynomial multiplication in lattice-based cryptography is proposed. The polynomial coefficients are stored in a ping-pong structure to improve the bandwidth. By eliminating pre-scale operations, 10.5% of modular multiplication operations and 16.7% of storage space are saved. The structure based on look-up table shift register and three-input adder is adopted to reduce the logical resource occupation. The pipeline structure with optional stages is designed to make the butterfly module in polynomial multiplication meet the timing requirements of different cryptographic hardware systems. The evaluation results show that the maximum frequency of low-area, balanced and high-performance implementations of the optimized butterfly unit can reach 150, 250 and 350 MHz, respectively. Compared with the existing implementation technologies, the optimized hardware implementation can achieve higher operating frequency with a smaller circuit area, which improves the efficiency of polynomial multiplication module by 22.8%. 

Different from the traditional method, the mathematical optimization model is established with Gauss principle to handle the singular problems to deal with the singular problems. Traditional optimization method and intelligent optimization method (particle swarm optimization algorithm, PSO) are combined to solve the above optimization problem, which can fully utilize the fast convergence of the traditional optimization method and the characteristic of global searching of the intelligent algorithm. The numerical example is simulated by Lagrangian formulation, null space method and Gauss optimization method respectively. The simulation results show that Gauss optimization method has higher computational accuracy, keeps the stability of the numerical calculation and would not lead to simulation failure due to the sudden changes of the system degree of freedom, which validates the effectiveness and universality of the proposed method.

 In order to study the overnight sleep condition and analyze each stage of the sleep process, polysomnography (PSG) and actigraphy were used to collect the ECG signal and body motion data. The features of ECG signal and heart rate variability (HRV) were extracted and used as the characteristic parameters of the data. In order to improve the recognition rate and prevent over-fitting, the data were divided into training set and test set, and an improved BP neural network model with genetic algorithm was designed to train and predict the samples. The results show that the improved BP neural network can effectively identify the test samples, and the comprehensive recognition accuracy is 86.29%. Wearable devices that detect both ECG and body motion signals with sleep stage classifying algorithms, can be used for family sleep monitoring and as a primary screen method for sleep disorders.

To solve the resource-lack problem of Chinese word concreteness, this paper designs and implements an automatic method to construct Chinese concreteness lexicon. By making full use of the existing resource of English word concreteness, it builds up a large-scale Chinese concreteness lexicon based on pretrained word embeddings and an MLP concreteness regression model. In addition, it proposes the concreteness inference tasks on the word level and on the sentence level, and manually constructs the corresponding datasets for evaluation the performance of the Chinese concreteness lexicon on these tasks. Experimental results show that the constructed concreteness lexicon can perform the two inference tasks effectively.

The interannual change rate of the normalized vegetation index (NDVI) and its correlation with climate factors were compared under different permafrost degradation and vegetation types. The results indicated that NDVI of the coniferous forest accelerated, which was positively correlated with the temperature and negatively with the precipitation in the growing season. As the permafrost active layer deepens, the rate of increase in NDVI of coniferous forests gradually decreases from north to south. The grassland NDVI accelerated in non-permafrost regions, which was positively correlated with growing season precipitation. There is a clear difference between the response of mixed forests to climate in permafrost and non-permafrost regions. In permafrost regions, mixed forest NDVI is positively correlated with growing season temperature and negatively correlated with growing season precipitation. As the permafrost active layer deepens, the correlation coefficient between the NDVI of mixed forest and growing season temperature shifts from positive to negative, and the correlation coefficient with growing season precipitation shifts from negative to positive. This may be related to the different water supply caused by different active layer thickness. The results imply that under the coupling effects of climate and permafrost, climate warming will lead to gradual northward shifting of coniferous forests and mixed forests, and occupation by grasslands in non-permafrost regions.

Based on the basic database of resources and environment, an index system was constructed to evaluate the ecological sensitivity of land resources in China. We revealed its spatial distribution characteristics among different land use types and different agricultural areas. The main conclusions are as follows. 1) There are significant spatial differences in ecological sensitivity of land resources in China. Highly ecological sensitive areas are concentrated in four regions: the northern arid/semi-arid deserts with land desertification sensitivity, Loess Plateau with soil-erosion sensitivity, the southern hills with soil-erosion sensitivity, and the karst region in the southwest which is both sensitive to rocky desertification and soil erosion. 2) The ecological sensitivity of cultivated land, forests and grassland is significantly different. The sensitivity of cultivated land is lower than the others. Dry land in the north is more sensitive than paddy field in the south; forests are more sensitive to soil erosion and rocky desertification, mainly in the Greater Khingan Range-Changbai Mountains; grassland is highly sensitive to land desertification, mainly refers to the low-cover grassland in eastern Inner Mongolia. 3) The ecological sensitivity of land resources in the nine agricultural areas is generally high in the north and low in the south. According to the characteristics of land ecological sensitivity in different agricultural areas, various land use measures and ecological protection strategies should be implemented.

In order to solve the problem that convolutional neural network is large and the accuracy loss of the model pruning method is relatively serious, a fine-tuning optimization method for model pruning is proposed. The global information of the original convolutional neural network model is introduced to the post-prune model to make it store the original model information which improves the accuracy of the model after pruning. Experimental results show that for the image classification tasks and target detection tasks, proposed fine-tuning optimization method can obtain greater compression ratio and smaller model accuracy loss. 

Based on the calculation method of First-Principles in DFT, the authors carry out the study of the properties of H2O molecules adsorbed on the KDP(100) surface. By topological analysis of the electron density. combining analysis methods such as Bader charge, electron density, electron density difference, electron localization function and other parameters, it is found that the best adsorption site for H2O molecules on KDP(100) surface is H-K bridge site where adsorption energy is –0.809 eV, indicating that the KDP(100) surface can absorb H2O molecules spontaneously. The oxygen atoms in H2O molecules form strong hydrogen bonds O—H...O_w involving covalent effect, with bond energy –18.88 kcal/mol.

Potential forecast models have been developed for air pollution of summertime (Apr.–Aug.) ozone and wintertime (Nov.–Feb.) PM_2.5 in Chengdu using the multiple linear regression (MLR), back-propagation (BP) neural network (NN) and random forest (RF) algorithms. The key predicting factors for each of the models are selected from various potential factors that may impact the spatiotemporal distribution of pollutions. The models are trained and established with 2016–2018 datasets and evaluated with a data-withheld method and further with independent 2019 dataset. The results show that the MLR, NN and RF models are all capable to accurately predict O₃ and PM_2.5 pollution potentials in short lead-time (1–3 days) in Chengdu. The models are also found having quite stable performances in medium- and long-term (7–15 days lead time) forecasts. Among the three models, the MLR model performs the best in prediction of O₃, while RF model performs the best for PM_2.5.

Wafer-to-wafer bond technology has breakthrough semiconductor manufacturing from 2D to 3D, but the bonded wafer brings more locating and patterning rules, it is too complex to layout the frame cells by traditional floor plan arrangement. This article provides a new floor plan arrangement method in face-to-face bonding product. It could setup all floor plans at same time only by flipping the motherboard. The new method is introduced. Final result with new method’s benefit is shown based on actually bonding product taping out procedure.

By applying the probabilistic benefit-cost analysis with Monte Carlo simulation, this study reveals the energy- and life-cycle climate and pollution-related health impacts, and various social costs in China of gasoline-to-electric vehicle (GV-to-EV) substitutions. Key factors influencing these estimates are also elicited. Results indicate that a GV-to-EV substitution can annually achieve benefits of 324 yuan on climate, 343 yuan on health, and 4315 yuan on energy savings in the energy cycle under current technical conditions and the dynamics of power grid development. However, these benefits are offset by the incremental manufacturing cost (16000 yuan/car-year), and the social welfare improvement over the life cycle is negative in the short run. This study highlights the importance of prioritizing GV-to-EV substitutions only in areas with high health, climate, and energy-saving benefits.

 In order to explore the main controlling factors of uplift deformation of Longmenshan structural belt, based on the differences in the properties of lower crust material layer between the east and west sides of Longmenshan structural belt, three groups of PFC2D discrete element numerical simulation are carried out to realize quantitative analysis. The experimental deformation results and the model particle motion vector map show that under the condition of no obvious difference in the material properties of the lower crust, the existence of plate collision and compression stress and crustal thickness difference will not form a huge topographic elevation difference in the Longmenshan structural belt. When there are obvious differences in the viscosity coefficient of the lower crust, the relative value of the particle movement rate of the weak lower crust material layer is 1.5?2.94 m/s, and the average movement rate is 1.62 m/s, which is about 54 times of the average movement rate of the particles of the hard lower crust layer. Uplift deformation occurs in the middle of the model (Longmenshan structural belt), with a vertical influence range of 94.74% and a uplift amplitude of 19.85%. The particles of the middle crust and upper crust overlying the weak lower crust have a large upward velocity component, and the upward trend of the material layer of the upper crust is obvious. There is a 20 km thickness difference between Bayankala block and the crust of Sichuan Basin, which increases the uplift amplitude of Longmenshan structural belt from 14.79% to 19.85%. Based on the comprehensive analysis of three discrete element simulation experiments, it is concluded that the viscosity difference between the material layer of the lower crust of Bayan Kara block and the material layer of the underground block of Sichuan Basin is the most key control factor for the vertical uplift deformation of Longmenshan structural belt. On the premise that there are obvious differences in the viscosity structure of the lower crust, the crustal thickness differences between the Bayan Kara block and the Sichuan Basin significantly promote the vertical thrust uplift amplitude of the Longmenshan structural belt. 

Taking rice SUPERWOMAN1 (SPW1)/OsMADS16 gene as an example, by optimizing the preparation of hybridization probes, material immobilization, dissociation, permeabilization and color development, the SPW1/OsMADS16 gene expression patterns with high specificity but low backgrounds were obtained. The procedure of whole mount in situ hybridization (WISH) technology for rice is simple and inexpensive, and it can be performed simultaneously with several species in the centrifuge tube. The main experimental steps include probe preparation, material immobilization and dissociation, hybridization reaction, post-hybridization treatment and detection and color development. The proposed method provides a technological basement for a protentional high throughput detection of spatiotemporal gene expression patterns and functional analysis in rice.

Aiming at the characteristics of complex entities in weapons and equipment field, a complex named entity recognition method is proposed which integrates multi-features and mounts the domain knowledge of weapons and equipment. First, we use the BERT model to pre-train on the weapon equipment field data to obtain the data vector, and use the Word2Vec model to learn context features of Zhengma, Wubi, Pinyin, and strokes to obtain the feature vector. Then the data vector and the feature vector are fused, the Bi-LSTM model is used for encoding, and the CRF decoder is used to obtain the tag sequence. Finally, the detection of complex entities on the label sequence is triggered to complete the recognition of complex named entities. In the experiments, we use the data collected from Global Military Network as the corpus, and analyze the recognition effect of different feature combinations and neural network models. A calculation method suitable for evaluating the recognition results of complex named entities is also proposed. The experimental results show that the F1-value of the proposed method for recognizing complex named entities of weapons and equipment with domain knowledge and fusion of multifeatures reaches 95.37%, which outperforms the existing methods.

Chaohuperleidus primus has yet to be completely described due to the limit of the original material which were collected from Upper Member of the Nanlinghu Formation (Spathian, Olenekian, Early Triassic) at Majiashan quarry, Chaohu City, Anhui Province. The taxon was redescribed in detail by adding three complete specimens from the type horizons. The generic diagnosis of Chaohuperleidus was revised mainly based on newly recognized anatomical information, of which the following characters ‘the fused parietal and dermopterotic with anterior, middle and posterior pit-lines; the operculum and suboperculum of nearly equal in height with the latter having a large rounded anterior dorsal process’ were possible apomorphies of Chaohuperleidus. The previous taxonomic assignment of the Chaohuperleidus was confirmed, which was similar to the Ladinian (Middle Triassic) genus Perleidus but differed the latter in having more suborbitals, branchiostegal rays and epaxial fin rays besides its possible apomorphies. The skull pattern of the Chaohuperleridus primus was exhaustively reconstructed and was anatomically compared with the Early Triassic taxa which were wrongly classified into the Perleidiformes and some newly described stem neopterygians. The result herein will provide new anatomical evidences for the phylogeny analysis of the stem neopterygians that is open to discuss.

The purpose of this study is to explore the potential changes in the relationship between dynamic light and photosynthesis in high CO₂ environment, so as to provide scientific support for coping with global changes. We characterized dynamic light environment at two heights within rice canopy and dynamic photosynthetic response of leaves at each height (about 45 cm or 90 cm high from the soil of paddy field) in an open canopy and within an open top chamber (OTC) with a high CO₂ of 600 μmol/mol air. The result showed that the averaged photosynthetic photon flux density incident at 90 cm height above the paddy field was 3–4 times higher than that at 45 cm height, but the coefficient of temporal variation in light received at 90 cm height was 55%–60% smaller than that of the lower layer. Flag leaves at 90 cm height showed higher saturated photosynthetic rate and more postillumination CO₂ fixation, while leaves at 45 cm height showed higher photosynthetic rate under low light. In the high CO₂ OTC, light attenuation percentage in canopy tended to increase, compared with leaves outside the OTC; the difference between assimilation rates at the two heights also increased within the OTC. These results suggested that not only the steady-state photosynthetic rate, but also the dynamic photosynthesis in rice leaves may have altered in different ways at different canopy heights, which is to be further influenced by CO₂ environment. 

A LSTM (long-short term memory) model is applied to the prediction of the Nino3.4 index, and the spring prediction barrier (SPB) issue has been further investigated in the LSTM model. The results show that the model can predict the trend of the Nino3.4 index well, yet revealing different performance in different El Nino events. For the 1997/1998 El Nino and 2015/2016 El Nino, which are strong EP El Nino events, the model performes well on the prediction of Nino3.4 index trend and peaks, and anomaly correlation coefficient (ACC) reaches more than 0.93. But for the weak CP El Nino events, e.g. the 1991/1992 El Nino and 2002/2003 El Nino, it shows relatively poor performance on the prediction of the peak. In the growing period, the maximum season growth rate of prediction error are in AMJ quarter, which indicates obvious SPB phenomenon. However, in the decaying period, the maximum have similar distribution in the same type of events: for the weak CP El Nino events, the maximum are in AMJ quarter, indicating obvious SPB phenomenon; for strong EP El Nino events, the maximum are in other quarter, indicating that there is no SPB phenomenon. The differences in the performance among individuals may be related to the development characteristics of the event itself (such as event type and intensity).

The nonlinear effect of power amplifier causes the strong nonlinear self-interference signal in the co-time co-frequency full duplex communication system, which reduces the communication performance of the system. A joint self-interference suppression method based on regularization is proposed. Taking into account both the multipath channel and the nonlinear characteristics of power amplifier, the proposed scheme can eliminate the linear and nonlinear self-interference signals, and alleviate the numerical instability of the traditional algorithms. To analyze the performance of the method, a simulation platform is built. Numerical simulation results show that the proposed scheme has higher gain than the traditional linear and nonlinear cancellation schemes.

Based on Byrnes’ definition and theory of critical thinking, through the research of approximately 1000 college and adult subjects, the authors have determined the structure of critical thinking of Chinese people and compiled the corresponding scale. In order to verify Byrnes’s theory of critical thinking structure, the authors first interviewed 40 subjects and generated a question bank for measuring critical thinking. Then, a preliminary scale was compiled. Second, an exploratory factor analysis was conducted on the data of 284 college students and the three dimensions of Chinese critical thinking — analytic ability, open-minded to criticism, and effort to use critical thinking, were obtained. Third, the authors conducted a confirmatory factor analysis on the data of 168 subjects, which confirmed the fit of the three-dimensional model. Finally, the data of 586 subjects were analyzed to further prove the reliability and validity of the scale. The result shows that the scale to measure the critical thinking of Chinese people conform to the theoretical constructs of Byrnes and colleagues. This scale has important theoretical and practical significance for future research in related fields.