The current deep learning studies on remote sensing mainly focused on deep learning algorithms rather than deep learning datasets. This study proposes a method of dataset consistency assessment based on deep learning, in which the similarity among various types of ships from different sources (such as satellite remote sensing, 3D modeling, and web crawler) is evaluated and then used to characterize the consistency of the ship dataset. The results show that when the consistency of the dataset is the highest, the consistency by the proposed method is 1. When the consistency of the datasets is gradient, the consistency also changes. Images with similar data sources can be considered as same class, and images with greatly differences cannot be merged. Thus, the proposed method can assess the dataset consistency properly, and provide a suggestion to build an image dataset for deep learning training. 

Evaluating the schemes and characteristics of models that are applied in studying the mechanism of how climate change would impact air pollution is key to providing a better understanding of the current studies and supporting modeling research in the future. This study reviews existing modeling studies on the topic “climate change impact on air pollution” based on literature investigation. Three types of modeling studies at different spatial scales – global and regional – are identified, and the characteristics and applicability of different research methods are compared. Furthermore, using data from the CMIP5 climate model intercomparison project and the atmospheric chemical transport model WRF-Chem, the impacts of future changes in meteorology and pollutant emissions in the context of climate change on near-surface summertime O₃ concentration in China in 2050 are studied, with the Beijing-Tianjin-Hebei region, Yangtze River Delta region, and Pearl River Delta region selected as representatives for analysis. Results show that under the RCP8.5 climate change and emission pathway, both meteorological and emission changes will impose significant effects on summer ozone concentrations in China by the middle of this century. In most regions of China, changes in emissions would have a significant influence on ozone concentration, but the meteorological field near the East China Sea would also significantly affect future ozone pollution. In addition, a certain degree of interaction between the two factors exists.

By fully utilizing the features of poetry text and related knowledge, this paper proposes a model SACAP for sentiment analysis of Chinese ancient poems integrating explicit knowledge and implicit knowledge. On the one hand, the model extracts deep semantic features from poetry text, and on the other hand, on the basis of constructing a knowledge base of Chinese ancient poetry, it designs a multidimensional attention mechanism to extract features from Chinese ancient poetry knowledge. The sentiment of Chinese ancient poetry can be determined by taking both features into consideration. The experimental results show that the proposed model performs better than existing models and the explicit knowledge plays much more important role, compared with implicit knowledge.

A reanalysis of marine adaptation of Qianosuchus mixtus is given by observing and making measurements to a new specimen. The marine adaptation in non-avian tetrapods follows a pattern of five steps (M1-M5). Comparation of the statistics of Qianosuchus mixtus and other clades leads to the result that Qianosuchus mixtus is placed in M2-3 and not considered to be completely marine. Observation and analysis of the morphology of the new specimen shows that its limbs do not exhibit characters of marine adaptation, yet its body shape and tail are morphologically adapted to undulatory or subundulatory swimming, similar to extant crocodiles and marine iguanas. Comparison with extant crocodiles and other clades suggest that despite having a specialized tail, Qianosuchus mixtus is not capable of swimming at high speed or for long distances. With the information above and reference to the habits of tetrapods in the Panxian Fauna, it is assumed that Qianosuchus mixtus possibly inhabits an intraplatform with islands, and that Qianosuchus mixtus and other upper trophic level predators exhibit complex ecology.

When using deep learning models to automatically generate radiology reports, due to the extreme imbalance of data, it is difficult for current models to identify abnormal regional features, which leads to misjudgment and missed judgment of the disease. In order to improve the model’s ability to identify diseases and improve the quality of reports, the authors use a multi-scale feature parsing Transformer (MFPT) model to generate radiology reports. Among them, a key feature enhanced attention (KFEA) module is constructed to strengthen the utilization of key features. A multi-modal feature fusion (MFF) module is designed to promote the feature fusion of semantic features and visual features and alleviate the impact caused by feature differences. This paper explores the role of stage-aware (SA) module in optimizing primary features in radiology reporting tasks. Finally, compared with the current mainstream models on the popular radiology report dataset IU X-Ray, the results show that the proposed model has achieved the current best effect.

In order to enhance the robustness of the visual question answering model, a bias reduction method is proposed. Based on this, the influence of language and visual information on bias effect is explored. Furthermore, two bias learning branches are constructed to capture the language bias, and the bias caused by both language and images. Then, more robust prediction results are obtained by using the bias reduction method. Finally, based on the difference in prediction probabilities between standard visual question answering and bias branches, samples are dynamically weighted, allowing the model to adjust learning levels for samples with different levels of bias. Experiments on VQA-CP v2.0 and other data sets demonstrate the effectiveness of the proposed method and alleviate the influence of bias on the model.

This paper proposes a deep learning-based time-domain denoising method for micro-earthquake recordings by combining a convolutional neural network (CNN) and a bidirectional long short-term memory network (BiLSTM). Based on micro-earthquake observation data from Zigong and Neijiang areas of Sichuan, the structural model and focal mechanism of the region are used to generate a synthetic noise-free dataset by numerical modeling, which is then combined with observed micro-earthquake noise to create a synthetic noisy dataset. A high-performance and stable denoising model is obtained through training of the deep learning network, demonstrating excellent generalization performance on the validation set. Compared with traditional methods, the proposed method demonstrates excellent denoising performance and better preserves the detailed characteristics of both the waveform and the spectrum of the noise-free signal. Application to micro-earthquake observation data of Zigong and Neijiang areas demonstrates the model’s strong denoising performance and generalization ability on real-world data.

Taking the Guanlan River in Shenzhen as an example, this study investigated the pollution level, spatiotemporal distribution, sources, and ecological risks of 20 phthalate esters (PAEs) in river water. It was found that in 2021, the concentrations of the 20 PAEs (Σ₂₀PAEs) ranged from 30.29–9755.87 ng/L with an average of 2550.73 ng/L in the dry season (DS) and 359.87–27247.01 ng/L with an average of 5262.87 ng/L in the wet season (WS), respectively, being higher in the WS than those in the DS. The Σ₂₀PAEs in the water of the Guanlan River were significantly lower in the main stream than those in the tributaries (p < 0.05). Moreover, their pollution was more serious downstream of wastewater treatment plants (WWTPs) than upstream. The contents of bis(2-ethylhexyl) phthalate (DEHP) and di-isobutyl phthalate (DIBP) were the highest in the DS and WS, respectively. The proportion of DEHP was significantly higher downstream of WWTPs than upstream (p < 0.05). Source analysis indicated that PAEs in the Guanlan River originated from diverse sources such as plastics, cosmetics, and personal care products. The ecological risk assessment revealed that PAEs posed the greatest risk to algae. The risk quotients (RQs) of most samples were greater than 1, indicating a high ecological risk of PAEs in the Guanlan River. Among them, DEHP should be given special attention.

Based on the ambient air quality monitoring data of 7 cities in Tibet from 2017 to 2023, the spatial and temporal distribution characteristics of PM₁₀ and PM_2.5 in ambient air in Tibet were analyzed by correlation analysis and GIS Kriging interpolation. The results showed that in Tibet region, the days with PM₁₀ and PM_2.5 as the primary pollutants were characterized by good ambient air quality, mainly concentrated in January–March and October–December. The variation trend of PM₁₀ and PM_2.5 concentration was consistent. The average annual concentration in the whole region reached the first-level standard and showed a downward trend, and the seasonal average concentration showed the characteristics of winter>spring>autumn>summer. The average monthly concentration began to decrease from May, reaching the lowest in August and the highest in December. The hourly concentration showed bimodal characteristics, and the peak appeared from 09:00 to 12:00 in the morning and from 21:00 to 01:00 in the next day. The concentrations of PM₁₀ and PM_2.5 in Nyingchi City in southeastern Tibet were significantly lower than those in other cities, while the concentrations of PM₁₀ and PM_2.5 in Nagqu City in northern Tibet were significantly higher than those in other cities. According to the analysis of Sugimoto model, in the exceedance days with particulate matter as the primary pollutant, dust weather had a particularly significant impact on the concentration of PM₁₀ in the atmospheric environment of Tibet, the average content of dust components in PM₁₀ was 87.9%. 

Based on the monitoring data of 18 river sections, this study investigated the spatiotemporal distribution, sources and ecological risks of 8 heavy metal ions in the upper reaches of Wujiang River, and analyzed the influence of dissolved organic matter (DOM) on the migration and distribution of heavy metals. The results showed that the overall content of heavy metals in the upper reaches of Wujiang River was low with significant seasonal variations, and the average concentrations of Cr, Fe, Co, Ni, Cu, As, Cd and Sb were 0.71, 19.20, 0.17, 1.09, 0.73, 0.65, 0.06 and 1.22 μg/L, respectively, which were all lower than the level of Class III of the Environmental Quality Standard for Surface Water. Principal component analysis revealed a homology among Cr, Fe, Co, Ni, Cd and Sb, primarily sourced from natural origins. Cu was additionally affected by industrial activities including electroplating, printing and dyeing, while as was partially originated from the application of pesticides and fertilizers. Heavy metals showed positive correlations with DOM and its fluorescent components C1–C4, indicating that DOM could promote the migration and enrichment of heavy metals in water. Cr, Fe, Co, Ni and Sb showed significant positive correlation with DOM characteristic parameters SUVA₂₅₄ and BIX, further confirming that DOM with higher aromaticity and biological reactivity showed higher metal affinities. The evaluation results of single factor and Nemerow comprehensive pollution index suggested that the pollution level of heavy metals in the upper reaches of the Wujiang River retained relatively low, rendering it an ideal source of drinking water. However, the potential ecological risks of heavy metals under the influence of DOM need more concern. 

Making full use of the economic properties of carbon sinks and the carbon trading market mechanism, scientific development of mangrove carbon sinks and the realization of emission reduction trading can bring additional economic benefits, and play a positive role in promoting the protection and restoration of mangroves. This article systematically summarizes the overview and the development technology of mangrove carbon sink, as well as the current status of carbon trading and its problems. The corresponding suggestions and countermeasures included as follows. 1) Strengthen basic research and promote the sustainable implementation of the development methodology of mangrove carbon sink projects in China. 2) Utilize the carbon inclusive market mechanism to expand the development path of mangrove carbon sinks. 3) Explore mangrove carbon sink trading and offsetting methods to improve mangrove carbon sink rate of return. 4) Encourage social capital to enter the development and trading of mangrove carbon sinks, and strengthen the protection and restoration of mangroves.

Based on Convolutional Long-Short Term Memory Neural Network, the authors proposed a deep learning method PressureConvLSTM to extract features during walking in both spatial and temporal dimensions. Classification based on plantar pressure of anterior cruciate ligament deficiency (ACLD) was applied to distinguish walking gait information. Experiment results combined with clinical data showed that PressureConvLSTM model obtained 95% test accuracy when diagnosing ACLD, which was well performed over other traditional deep learning models.

The basic concepts and core issues related to robot learning are introduced and discussed, and the relevant researches are summarized and analyzed. Through comparing the relevant methods and recent progress, the authors classify the methods of robot learning into four categories based on data types and learning methods, namely reinforcement learning approach, imitation learning approach, transfer learning approach and developmental learning approach. Finally, current challenges and future trends in robot learning are listed.

In order to accurately grasp the emission characteristics of anthropogenic air pollutants in Kunming, this study utilizes statistical data from the statistical yearbook and various municipal departments, integrating enterprise surveys, field sampling, and on-site interviews and investigations to establish an anthropogenic emission inventory for Kunming in 2018. The source spectrum data for volatile organic compounds (VOCs) in Kunming were obtained through emission outlet testing at key enterprises and literature research. A detailed list of VOCs sub-species in Kunming was compiled, and their ozone formation potential was calculated. The results showed that the emissions of sulfur dioxide (SO₂), nitrogen oxides (NO_x), carbon monoxide (CO), VOCs, ammonia (NH₃), particulate matter (PM₁₀), fine particulate matter (PM_2.5), black carbon (BC) and organic carbon (OC) in Kunming during 2018 were 13476.92 tons, 53327.85 tons, 397383.83 tons, 55514.73 tons, 20465.41 tons, 75473.99 tons, 29405.57 tons, 1947.53 tons, 4405.39 tons, respectively. Among these, the primary emission sources for NO_x were mobile sources (50.7%), NH₃ were agricultural sources (88.5%), PM₁₀ were dust sources (44.1%) and process sources (43.1%). The main emission sources of CO, VOCs and PM_2.5 were process sources, which accounted for 68.2%, 41.7%, and 51.2% of the emissions of different pollutants, respectively. The primary emission sources for SO₂, BC and OC were stationary combustion sources of fossil fuel, with emission shares of 53.0%, 45.0% and 35.9%, respectively. Pollutants were mainly concentrated in the 5 districts of the main city as well as in Anning City. Within the 5 districts of the main city, pollutants were distributed outward from the center along Youth Road and Renmin Middle Road, with relatively few pollutants found in Chenggong District. SO₂, BC, and OC were mainly distributed by high-value point sources, NO_x, CO, VOCs, and PM_2.5 were distributed by a combination of line and point sources. PM₁₀ presented a spatial distribution characterized by a combination of point and surface sources. NH₃ showed a significant spatial distribution characteristics of surface sources. VOCs sub-species inventory emissions were dominated by aromatic hydrocarbons and alkanes, with the main sources being vehicle emissions and architectural coatings, as well as the industrial solvent. Ozone formation potentials (OFP) for aromatic hydrocarbons accounted for up to 49.9%, with species such as (m- and p-) xylene, toluene, and ethylene comprising a relatively high proportion of the VOCs species. 

This article aims to translate and revise the Digital Stress Scale (DSS) developed by Hall et al. (2021) and validate the revised Chinese DSS’s reliability and validity among Chinese college students. Structured interviews regarding social media use and digital stress were conducted with 15 Chinese college students in sample 1. Four other samples of Chinese college students (n=87, n=100, n=300, n=239) were recruited online by convenience sampling. Item analysis, confirmatory factor analysis, construct validity, criterion-related validity and reliability analysis were conducted. Depression-Anxiety-Stress scale, life satisfaction, UCLA loneliness scale, perceived social support and Bergen Social Media Addiction Scale were used to assess criterion-related validity. Test-retest validity was assessed among 156 college students in sample five two weeks after baseline. The revised Chinese Digital Stress Scale (RC-DSS) consists of 31 items, including six dimensions (availability stress, approval anxiety, social comparison, fear of missing out, connection overload and online vigilance). Discrimination analysis and item analysis showed good discriminability. The six-dimension structure of the scale was confirmed by confirmatory factor analysis (χ2/df=2.82, GFI=0.80, NFI=0.93, TLI=0.95, CFI=0.96, RMSEA=0.08). Digital stress was significantly and moderately associated with social media addiction, stress, depression, anxiety and loneliness (rs=0.41–0.61, ps<0.01), and was negatively associated with social support and life satisfaction (r = −0.24, p<0.01; r = −0.15, p<0.05). The Cronbach’s alpha of the scale was 0.94 and its two-week test-retest reliability was 0.73 (p<0.01). The RC-DSS is a reliable and valid instrument to assess digital stress among Chinese college students.

In order to reveal the trend of water chemistry change in the Yarlung Tsangpo River (Yajiang River) under the background of climate change and provide scientific and technological support for water resources and water environment management in the basin, based on the study of hydrochemical characteristics of the Yajiang River in 2016, 2017 and 2018, combined with the research results of existing research teams on hydrochemistry, this paper studies the trend of 11 hydrochemical components change of the upper, middle and lower reaches of the Yajiang River over the past 60 years by comprehensive use of linear tendency estimation, climate change model output and BP neural network model. The results show that the annual average temperature in the upper, middle and lower reaches of the Yajiang River Basin has been increasing obviously in the past 60 years. The average temperature warming rate was 0.38°C/10a. The precipitation in the Yajiang River Basin fluctuated obviously and showed an overall rising trend, with a rising rate of 7.34 mm/10a. pH value of the water in Yajiang River was weakly alkaline and showd an upward trend. TDS was higher than the average level of the world river (120 mg/L) and showed a trend of gradual increase. Based on the climate change model RCP4.5 scenario, the artificial neural network prediction shows that the TDS flux in the upper, middle and lower reaches of the Yajiang River Basin presents a gradually increasing trend, and the downstream will have a certain impact on the production and life of the downstream residents. 

Network analysis, element correlation analysis and phylogenetic analysis are applied in the visualization methods study of mineral crystal chemistry data. Taking P-bearing minerals as an example, force-directed network and bipartite network diagram of mineral composition and genesis, phylogenetic tree of mineral crystal characteristics and correlation heat maps of mineral component elements are drawn. These methods also take into account the spatial and temporal distribution, evolutionary diversity and physical and chemical properties of minerals. The use of these visual analysis methods is helpful to explore the evolution of the Earth’s environment using mineralogical records through deep time and understand its evolutionary process and driving mechanism.

This paper proposed a bi-directional attention based text-keyword matching model for law recommendation (BiAKLaw). In this model, BERT is utilized as a basic matching model, bi-directional attention mechanism is implemented to extract token-level alignment features and keyword-level differential features, and these features are fused with keyword attentive semantic representations for a better matching effect. The experimental results on the traffic accident and intentional injury datasets demonstrate that, compared with BERT, the proposed model increases F1 evaluation metric by 3.74% and 3.43% respectively.

Most existing neural network-based approaches for query suggestion use solely query sequences in query logs as training data. However, these methods cannot fully mine and infer all kinds of semantic relationships among words or concepts from query sequences because queries in query sequences inherently suffer from a lack of syntactic relation, even a loss of semantics. To solve this problem, this paper proposes a new neural network model based on multi-source data augmentation through cross-attention (MDACA) for generating context-aware query suggestions. Proposed model adopts a Transformer-based encoder-decoder model that incorporates document-level semantics and global query suggestions into query-level information through cross-attention. The experimental results show that in contrast to the current suggestion models, the proposed model can generate context-aware query suggestions with higher relevance.

Based on information entropy, deviation coefficient of structural change rates and vector auto-regression model, taking Quzhou County as a typical case, this paper made a comprehensive study on the dynamic change characteristics and interactive relationship of industrial structure and land use structure of rural areas in Huang-Huai-Hai Plain. The results show that in recent years, the evolution of industrial structure in Quzhou County has experienced the process of primary industry dominating, balance of primary and secondary industries and secondary industry dominating. The great proportion of land for the primary industry leads to the extremely unbalanced land use structure, but the land use structure is developing to an equilibrium state. Industrial structure adjustment and land use structure adjustment are uncoordinated, and land use structure adjustment lags behind, but there is a long-term and stable relationship between them. Before 2009, the adjustment of industrial structure promoted the change of land use structure, but after that, the causality was opposite. In the later stage, the driving effect of land use structure adjustment on the subsequent optimization of land use structure and the upgrading of industrial structure was more lasting and significant. Over a period time in the future, optimizing the land use structure of the three industries by improving the land use efficiency of the primary industry, exploiting the land use potential of the secondary industry, and widening the land use scope of the tertiary industry is an effective way to promote the continuous upgrading of industrial structure in rural areas of Huang-Huai-Hai Plain.