Search

Complicated nano- and atomic-scale processes with sub-angstrom spatial resolution and millisecond time resolution visualized by in situ transmission electron microscopy (TEM) are often highly dynamical and time consuming to analyze and interpret. Here, we report how variational autoencoders (VAEs) can provide an artificial intelligence's interpretation of high-resolution in situ TEM data by condensing and deconvoluting complicated atomic-scale dynamics into a latent space with reduced dimensionality. We designed a VAE model with high latent dimensions capable of deconvoluting information from complex high-resolution TEM data. We demonstrate how this model, with high latent dimensions trained on atomically resolved TEM images of lead sulfide (PbS) nanocrystals, is able to capture movements and perturbations of periodic lattices in both simulated and real in situ TEM data. The VAE model shows the capability of detecting and deconvoluting dynamical nanoscale physical processes, such as the rotation of crystal lattices and intra-particle ripening during the annealing of semiconductor nanocrystals.

BackgroundCompared to commonly-used green space indicators from downward-facing satellite imagery, street view-based green space may capture different types of green space and represent how environments are perceived and experienced by people on the ground, which is important to elucidate the underlying mechanisms linking green space and health.ObjectivesThis study aimed to evaluate machine learning models that can classify the type of vegetation (i.e., tree, low-lying vegetation, grass) from street view images; and to investigate the associations between street green space and socioeconomic (SES) factors, in Los Angeles County, California.MethodsSES variables were obtained from the CalEnviroScreen3.0 dataset. Microsoft Bing Maps images in conjunction with deep learning were used to measure total and types of street view green space, which were compared to normalized difference vegetation index (NDVI) as commonly-used satellite-based green space measure. Generalized linear mixed model was used to examine associations between green space and census tract SES, adjusting for population density and rural/urban status.ResultsThe accuracy of the deep learning model was high with 92.5% mean intersection over union. NDVI were moderately correlated with total street view-based green space and tree, and weakly correlated with low-lying vegetation and grass. Total and three types of green space showed significant negative associations with neighborhood SES. The percentage of total green space decreased by 2.62 [95% confidence interval (CI): -3.02, -2.21, p < 0.001] with each interquartile range increase in CalEnviroScreen3.0 score. Disadvantaged communities contained approximately 5% less average street green space than other communities.ConclusionStreet view imagery coupled with deep learning approach can accurately and efficiently measure eye-level street green space and distinguish vegetation types. In Los Angeles County, disadvantaged communities had substantively less street green space. Governments and urban planners need to consider the type and visibility of street green space from pedestrian's perspective.

Abstract Background Though bictegravir/emtricitabine/tenofovir (BIC/FTC/TAF) have been regulatory approved and included in the National Reimbursement Drug List in China, due to the affordability concern, generic version of efavirenz + lamivudine + tenofovir (EFV + 3TC + TDF) is still recommended as the first-line therapy in the clinical guideline and widely used in clinical practice. The aim of the study is to assess the persistence with first-line BIC/TAF/TAF and EFV + 3TC + TDF in newly treated HIV-1 patients in the real-world setting in Hunan Province in China. Methods A retrospective analysis of the medical records of HIV patients initiating first-line antiretroviral therapy in the First Hospital of Changsha in January 1st, 2021-July 31st, 2022 was conducted. Persistence was assessed as the number of days on the therapy from the index until treatment discontinuation or end of data availability. Kaplan-Meier Curves and Cox Proportional Hazard models were used to evaluate the discontinuation rates. Subgroup analysis was performed excluding BIC/FTC/TAF patients with treatment discontinuation due to economic reason, and EFV + 3TC + TDF patients with a viral load > 500,000 copies/mL. Results A total of 310 eligible patients were included in the study, with 244 and 66 patients in the BIC/FTC/TAF group and EFV + 3TC + TDF group, respectively. Compared with EFV + 3TC + TDF patients, BIC/FTC/TAF patients were older, more living in the capital city currently, and had significantly higher total cholesterol and low-density level (all p 500,000 copies/mL, the analysis showed similar results (HR = 10.1, 95% CI = 1.2–84.1). 79.4% of the EFV + 3TC + TDF patients discontinued treatment due to clinical reasons, while 83.3% of the BIC/FTC/TAF patients discontinued treatment due to economic reasons. Conclusions Compared with BIC/FTC/TAF, EFV + TDF + 3TC patients were significantly more likely to discontinue the first-line treatment in Hunan Province in China.

Increasing access to undergraduate research is critical in efforts to retain students pursuing careers in STEM. Alternatives to traditional research positions, such as course based undergraduate research experiences (CUREs), have played important roles in engaging more undergraduates in research. However, these opportunities are only a subset of possible nontraditional research experiences which can enrich the undergraduate experience. In this study, we critically examined and improved the research-group-led undergraduate research program (GURP) to better meet the needs of undergraduates seeking research. Specifically, we investigated if the program was successful with a greater diversity of data sets, if assessments of learning were consistent across different data sets and teaching environments, and if the program promoted student engagement in research. We have found that this model is scalable, robust, and adaptable to different implementations while producing consistent and positive learning outcomes for students. Especially remarkable are students' increased self-identification as scientists and statistically significant gains in self-perceived competency across multiple domains of knowledge. This program model has shown promising results as a partially and fully online research experience for undergraduates and has benefited program alumni in their research careers. To assist research groups starting similar programs, we have created public data sets and instructional resources. We believe that GURP programs can work in a variety of situations and hope that they can become a tool to increase interest and build communities for young researchers.

The fluid distribution is considerably influenced by the reservoir homogeneity. This study analyzes carbonate reservoir heterogeneity and hydrocarbon accumulation processes in the Khasib reservoirs in AB oilfield in the Middle East to discuss factors controlling the distribution of fluid and the effects of this distribution on the productivity of single wells. The Khasib reservoirs in the AB oilfield show uniform vertical distributions, with a large difference in the longitudinal pore throats, and the upper part of the reservoir is better than the lower one. Because the hydrocarbon accumulation is multiphasic, the heterogeneity of crude oil is based on whether traps are formed in the early or late stage of source rock formation during accumulation. Moreover, the longitudinal physical differences in crude oil properties are mainly attributed to the longitudinal differences in the pore throats. The difference in the pore throat and physical properties of crude oil affords differences in the productivity of horizontal wells in different subzones.

Recently, two-dimensional magnetic materials (2DMMs) have become a focused research direction in a broad range of two-dimensional materials, due to their underlying significance in fundamental research, as well as in technologically relevant applications for future spintronics, magnonics, quantum information and data storage. The rich toolbox of 2DMMs and their diverse tunability have enabled the unprecedented research concerning the two-dimensional magnetic order down to single atomic layer of materials, much beyond the classical thin film magnetism, showing an extremely promising avenue towards electronics, magneto-optics and photonics. Among various degrees of freedoms, the interaction between spin and phonon (i.e., quanta of lattice vibration), hence the so-called spin-phonon coupling, serves as an important tuning knob to explore the two-dimensional magnetism, creates new types of quasiparticles and controls the magnetic order. This review overviews the latest developments in spin-phonon coupling investigations in 2DMMs. Various techniques utilizing spin-phonon coupling to investigate two-dimensional magnetism are discussed. Recent progress in tuning two-dimensional magnetic order based on spin-phonon coupling is also summarized, with a focus to understand the new functionalities. Furthermore, device developments and concepts based on spin-phonon coupling are briefly discussed. This review will provide our perspectives on the existing challenges and future directions in spin-phonon coupling research in 2DMMs and their functional devices.

During the joint extraction of entity and relationship from the operational management data of hydraulic engineering, complex sentences containing multiple triplets and overlapping entity relations often arise. However, traditional joint extraction models suffer from a single-feature representation approach, which hampers the effectiveness of entity relation extraction in complex sentences within hydraulic engineering datasets. To address this issue, this study proposes a multi-feature joint entity relation extraction method based on global context mechanism and graph convolutional neural networks. This method builds upon the Bidirectional Encoder Representations from Transformers (BERT) pre-trained model and utilizes a bidirectional gated recurrent unit (BiGRU) and global context mechanism (GCM) to supplement the contextual and global features of sentences. Subsequently, a graph convolutional network (GCN) based on syntactic dependencies is employed to learn inter-word dependency features, enhancing the model's knowledge representation capabilities for complex sentences. Experimental results demonstrate the effectiveness of the proposed model in the joint extraction task on hydraulic engineering datasets. The precision, recall, and F1-score are 86.5%, 84.1%, and 85.3%, respectively, all outperforming the baseline model. [ABSTRACT FROM AUTHOR]

High-energy exciton emission could allow single-component multi-colour display or white light-emitting diodes. However, the thermal relaxation of high-energy excitons is much faster than the photon emission of them, making them non-emissive. Here, we report quantum dots with light hole-heavy hole splitting exhibiting strong high-energy exciton electroluminescence from high-lying light holes, opening a gate for high-performance multi-colour light sources. The high-energy electroluminescence can reach 44.5% of the band-edge heavy-hole exciton emission at an electron flux density Φe of 0.71 × 1019 s−1 cm−2 − 600 times lower than the photon flux density Φp (4.3 × 1021 s−1 cm−2) required for the similar ratio. Our simulation and experimental results suggest that the oscillator strength of heavy holes reduces more than that of light holes under electric fields. We attribute this as the main reason for strong light-hole electroluminescence. We observe this phenomenon in both CdxZn1-xSe-ZnS and CdSe-CdS core-shell quantum dots exhibiting large light hole-heavy hole splittings. Previous observations of high-energy exciton electroluminescence rely on filling the ground state with excessive charges, Wang et al. report such observation from quantum dots at low charge concentration by increasing the relative oscillator strength of higher-lying light holes comparing to that of ground-state heavy holes. [ABSTRACT FROM AUTHOR]

Due to the frequent occurrence of extreme weather in recent years, accurate runoff prediction is crucial for the rational planning and management of water resources. Addressing the high uncertainty and multiple influencing factors in runoff prediction, this paper proposes a runoff prediction method driven by multi-source data. Based on multivariate observed data of runoff, water level, temperature, and precipitation, a Time2Vec-TCN-Transformer model is proposed for runoff prediction research and compared with LSTM, TCN, and TCN-Transformer models. The results show that the Time2Vec-TCN-Transformer model outperforms other models in metrics including MAE, RRMSE, MAPE, and NSE, demonstrating higher prediction accuracy and reliability. By effectively combining Time2Vec, TCN, and Transformer, the proposed model improves the MAPE for forecasting 1–4 days in the future by approximately 7% compared to the traditional LSTM model and 4% compared to the standalone TCN model, while maintaining NSE consistently between 0.9 and 1. This model can better capture the periodicity, long-term scale information, and relationships among multiple variables of runoff data, providing reliable predictive support for flood forecasting and water resources management. [ABSTRACT FROM AUTHOR]

The domain terms in power grid operation management documents are professional and complex. In the process of information extraction, excellent and applicable event detection methods are needed to extract event subjects. However, most of the current Chinese event detection methods use the word embedding technology to capture semantic representation, but it is difficult for these methods to capture the dependency between trigger words and other domain words in the same sentence. Based on the above situation, this paper proposes a novel hybrid representation architecture to represent the semantic and structural information of two characters and words. The model can capture rich semantic features through the semantic representation generated by the dependency parser. The experiment is based on the corpus of dispatching log, dispatching maintenance ticket, and dispatching plan. The results show that this method can significantly improve the performance of power grid dispatching text event detection.

To meet the demand of scalability and decision-making foresight of the traditional dispatching automation system in the new power system, a novel pre-dispatching method of new generation dispatching and control was proposed. First, the power grid digital twin was established in the description subsystem layer, which can reflect the state of power grid primary equipment, secondary equipment, and environment. Then, in the prediction subsystem level, the deep learning models were used to learn and predict future situation or accident risk of power grid operation in power grid digital twin. Finally, the feasibility of the proposed method was verified by the implementation example of East China Grid. The application results show that the pre-dispatching method improves the efficiency of system in dealing with the operation control problems of the new power system, which also provides a useful reference for comprehensive construction, popularization, and application of new generation power systems.