Hybrid Brain-Computer Interface Controlled Soft Robotic Glove for Stroke Rehabilitation.

Soft robotic glove controlled by a brain-computer interface (BCI) have demonstrated effectiveness in hand rehabilitation for stroke patients. Current systems mostly rely on static visual representations for patients to perform motor imagination (MI) tasks, resulting in lower BCI performance. Therefore, this study innovatively used MI and high-frequency steady-state visual evoked potential (SSVEP) to construct a friendly and natural hybrid BCI paradigm. Specifically, the stimulation interface sequentially presented decomposed action pictures of the left and right hands gripping a ball, with the pictures flashing at specific stimulation frequencies (left: 34 Hz, right: 35 Hz). Integrating soft robotic glove as feedback, we established a comprehensive "peripheral - central - peripheral" hand rehabilitation system to facilitate the hand rehabilitation of patients. Filter bank common spatial pattern (FBCSP) and filter bank canonical correlation analysis (FBCCA) algorithms were used to identify MI and SSVEP signals, respectively. Additionally, to fuse the features of these two signals, we proposed a novel fusion algorithm for improving the recognition accuracy of the system. The feasibility of the proposed system was validated through online experiments involving 12 healthy subjects and 9 stroke patients, achieving accuracy rates of 95.83 ± 6.83% and 63.33 ± 10.38%, respectively. The accuracy of MI and SSVEP in 12 healthy subjects reached 81.67 ± 15.63% and 95.14 ± 7.47%, both lower than the accuracy after fusion, these results confirmed the effectiveness of the proposed algorithm. The accuracy rate was more than 50% in both healthy subjects and patients, confirming the effectiveness of the proposed system.

Production of Oximes Directly from Sustainable Lignocellulose-Derived Aldehydes and Ammonia over HTS-1 Catalyst.

Oxime chemicals are the building blocks of many anticancer drugs and widely used in industry and laboratory. A simple but robust hierarchically porous zeolite (HTS-1) catalyst was prepared by hydrothermal methods and used for the preparation of vanillin oxime from vanillin in NH3 ⋅ H2 O/DIO (v/v 1/10) system. The results of the catalyst characterization showed that the larger pore size and more framework Ti were conducive to promote the transformation of the substrates. The conversion of vanillin and the yield of vanillin oxime were both higher than 99 % under optimized reaction conditions. It was found that the reaction proceeded by oxidation of NH3 to hydroxylamine (NH2 OH), and oximation of hydroxylamine with vanillin to obtain vanillin oxime, where the rate-controlling step was the hydroxylamine formation, and the apparent activation energy was 26.22 kJ/mol. The corresponding oximation products could also be obtained by extending this method to other compounds derived from lignin. Furthermore, the catalytic system was used directly to the conversion of birch biomass to obtain oxime products such as vanillin oxime, syringaldehyde oxime, and furfural oxime etc. This work might give insights into the sustainable production of N-containing high-value products from lignocellulose.

SVD-AE: An asymmetric autoencoder with SVD regularization for multivariate time series anomaly detection.

Anomaly detection in multivariate time series is of critical importance in many real-world applications, such as system maintenance and Internet monitoring. In this article, we propose a novel unsupervised framework called SVD-AE to conduct anomaly detection in multivariate time series. The core idea is to fuse the strengths of both SVD and autoencoder to fully capture complex normal patterns in multivariate time series. An asymmetric autoencoder architecture is proposed, where two encoders are used to capture features in time and variable dimensions and a shared decoder is used to generate reconstructions based on latent representations from both dimensions. A new regularization based on singular value decomposition theory is designed to force each encoder to learn features in the corresponding axis with mathematical supports delivered. A specific loss component is further proposed to align Fourier coefficients of inputs and reconstructions. It can preserve details of original inputs, leading to enhanced feature learning capability of the model. Extensive experiments on three real world datasets demonstrate the proposed algorithm can achieve better performance on multivariate time series anomaly detection tasks under highly unbalanced scenarios compared with baseline algorithms.

Modeling drug-induced mitochondrial toxicity with human primary cardiomyocytes.

Mitochondrial toxicity induced by therapeutic drugs is a major contributor for cardiotoxicity, posing a serious threat to pharmaceutical industries and patients' lives. However, mitochondrial toxicity testing is not incorporated into routine cardiac safety screening procedures. To accurately model native human cardiomyocytes, we comprehensively evaluated mitochondrial responses of adult human primary cardiomyocytes (hPCMs) to a nucleoside analog, remdesivir (RDV). Comparison of their response to human pluripotent stem cell-derived cardiomyocytes revealed that the latter utilized a mitophagy-based mitochondrial recovery response that was absent in hPCMs. Accordingly, action potential duration was elongated in hPCMs, reflecting clinical incidences of RDV-induced QT prolongation. In a screen for mitochondrial protectants, we identified mitochondrial ROS as a primary mediator of RDV-induced cardiotoxicity. Our study demonstrates the utility of hPCMs in the detection of clinically relevant cardiac toxicities, and offers a framework for hPCM-based high-throughput screening of cardioprotective agents.

Magnetic Prussian blue nanoshells are controllable anchored on the surface of molybdenum disulfide nanosheets for efficient separation of radioactive cesium from water.

The rapid development of nuclear energy in China has led to increased attention to the treatment of radioactive wastewaters. Herein, a novel magnetic adsorbent, magnetic Prussian blue­molybdenum disulfide (PB/Fe3O4/MoS2) nanocomposite, was prepared by a simple in-situ fixation of ferric oxide nanoparticles (Fe3O4 NPs) and Prussian Blue (PB) shell layers on the surface of molybdenum disulfide (MoS2) nanosheets carrier. The prepared PB/Fe3O4/MoS2 nanocomposites adsorbent displayed excellent fast magnetic separation and adsorption capacity of Cs+ (Qm = 80.51 mg/g) from water. The adsorption behavior of Cs+ by PB/Fe3O4/MoS2 conformed to Langmuir isothermal and second-order kinetic model, which belonged to chemical adsorption and endothermic reaction. The equilibrium adsorption capacity of PB/Fe3O4/MoS2 to Cs+ has reached 90 % in less than 110 min. Moreover, the adsorption properties of PB/Fe3O4/MoS2 remained good in the pH range of 2-7. Based on this, PB/Fe3O4/MoS2 complex was a fast and high selectivity adsorption material for Cs+, which was expected to be used in the practical treatment of cesium-containing radioactive wastewater.

Adaptive Transfer of Graph Neural Networks for Few-Shot Molecular Property Prediction.

Few-Shot Molecular Property Prediction (FSMPP) is an improtant task on drug discovery, which aims to learn transferable knowledge from base property prediction tasks with sufficient data for predicting novel properties with few labeled molecules. Its key challenge is how to alleviate the data scarcity issue of novel properties. Pretrained Graph Neural Network (GNN) based FSMPP methods effectively address the challenge by pre-training a GNN from large-scale self-supervised tasks and then finetuning it on base property prediction tasks to perform novel property prediction. However, in this paper, we find that the GNN finetuning step is not always effective, which even degrades the performance of pretrained GNN on some novel properties. This is because these molecule-property relationships among molecules change across different properties, which results in the finetuned GNN overfits to base properties and harms the transferability performance of pretrained GNN on novel properties. To address this issue, in this paper, we propose a novel Adaptive Transfer framework of GNN for FSMPP, called ATGNN, which transfers the knowledge of pretrained and finetuned GNNs in a task-adaptive manner to adapt novel properties. Specifically, we first regard the pretrained and finetuned GNNs as model priors of target-property GNN. Then, a task-adaptive weight prediction network is designed to leverage these priors to predict target GNN weights for novel properties. Finally, we combine our ATGNN framework with existing FSMPP methods for FSMPP. Extensive experiments on four real-world datasets, i.e., Tox21, SIDER, MUV, and ToxCast, show the effectiveness of our ATGNN framework.

Association of TyG index with hypertension in Chinese adults: the China Health Examination Collaborative Study (CHEC Study).

BACKGROUND AND OBJECTIVES: We aimed to investigate the association of triglyceride-glucose (TyG) index with hypertension and compare the discriminative power of the TyG index, lipid, glycemic parameters for hypertension using the China Health Examination Collaborative study (CHEC Study). METHODS AND STUDY DESIGN: Data were collected at Ningbo Mingzhou Hospital and Beijing physical examination center from the CHEC Study during 2014 and 2021. Participants with ≥2 medical check-up times were included. The TyG index is the logarithmized product of fasting triglyceride and glucose. Generalised estimation equation (GEE) model was used to evaluate the association between the TyG index, lipid parameters, glycemic parameters and hypertension. Receiver operating characteristic (ROC) analysis was performed to explore the predictive ability of TyG index on hypertension at different years of medical check-up. RESULTS: 112,902 participants with an average age of 42.8 years were recruited in the study, 36,839 participants developed hypertension over the 8-year period. GEE model analysis showed that the ORs with 95% CI of hypertension were 3.35 (3.15-3.57), 1.86 (1.76-1.95), 1.67 (1.58-1.78), 1.45 (1.33-1.58), 1.24 (1.19-1.29), 0.92 (0.86-0.99), and 1.90 (1.83-1.97) in the highest versus lowest quintiles of TyG index, TG/HDL-C ratio, TG, TC, LDL-C, HDL-C and FPG in model 2. The area under the ROC curve of the overall years of medical check-up was signifi-cantly higher than a particular year in predicting hypertension (AUC: 0.883, p < 0.05). CONCLUSIONS: TyG index is associated with hypertension and shows the superior discriminative ability for hypertension compared with lipid and glycemic parameters.

A review of the occurrence and degradation of biodegradable microplastics in soil environments.

The use of plastics for manufacturing of products and packaging has become ubiquitous. This is because plastics are cheap, pliable, and durable. However, these characteristics of plastics have also led to their disposal in landfill, where they persist. To overcome the environmental challenge posed by conventional plastics (CPs), biodegradable plastics (BDPs) are increasingly being used. However, BDPs form residual microplastics (MPs) at a rate that far exceeds that of CPs, and MPs have negative impacts on the soil environment. This review aimed to evaluate whether the move away from CPs to BDPs is having an overall positive impact on the environment considering the formation of MPs. Topics focused on in this review include the degradation of BDPs in the soil environment and the impacts of MPs originating from BDPs on soil physical and chemical properties, microbial communities, animals, and plants. The information collated in this review can provide scientific guidance for sustainable development of the BDPs industry.

UNIMEMnet: Learning long-term motion and appearance dynamics for video prediction with a unified memory network.

As a pixel-wise dense forecast task, video prediction is challenging due to its high computation complexity, dramatic future uncertainty, and extremely complicated spatial-temporal patterns. Many deep learning methods are proposed for the task, which bring up significant improvements. However, they focus on modeling short-term spatial-temporal dynamics and fail to sufficiently exploit long-term ones. As a result, the methods tend to deliver unsatisfactory performance for a long-term forecast requirement. In this article, we propose a novel unified memory network (UNIMEMnet) for long-term video prediction, which can effectively exploit long-term motion-appearance dynamics and unify the short-term spatial-temporal dynamics and long-term ones in an architecture. In the UNIMEMnet, a dual branch multi-scale memory module is carefully designed to extract and preserve long-term spatial-temporal patterns. In addition, a short-term spatial-temporal dynamics module and an alignment and fusion module are devised to capture and coordinate short-term motion-appearance dynamics with long-term ones from our designed memory module. Extensive experiments on five video prediction datasets from both synthetic and real-world scenarios are conducted, which validate the effectiveness and superiority of our proposed method UNIMEMnet over state-of-the-art methods.

An optimal medicinal and edible Chinese herbal formula attenuates particulate matter-induced lung injury through its anti-oxidative, anti-inflammatory and anti-apoptosis activities.

Objective: Identifying novel strategies to prevent particulate matter (PM)-induced lung injury is crucial for the reduction of the morbidity of chronic respiratory diseases. The combined intervention represented by herbal formulae for simultaneously targeting multiple pathological processes can provide a more beneficial effect than the single intervention. The aim of this paper is therefore to design a safe and effective medicinal and edible Chinese herbs (MECHs) formula against PM-induced lung injury. Methods: PM-induced oxidative stress, inflammatory response and apoptosis A549 cell model were used to screen anti-oxidant, anti-inflammatory and anti-apoptotic MECHs, respectively. A network pharmacology method was utilized to rationally design a novel herbal formula. Ultra performance liquid chromatography-mass spectrometer was utilized to assess the quality control of MECHs formula. The excretion of magnetic iron oxide nanospheres of the MECHs formula was estimated in zebrafish. The MECH formula against PM-induced lung injury was investigated with mice experiments. Results: Five selected herbs were rationally designed to form a new MECH formula, including Citri Exocarpium Rubrum (Juhong), Lablab Semen Album (Baibiandou), Atractylodis Macrocephalae Rhizoma (Baizhu), Mori Folium (Sangye) and Polygonati Odorati Rhizoma (Yuzhu). The formula effectively promoted the magnetic iron oxide nanospheres excretion in zebrafish. The mid/high dose formula significantly prevented PM-induced lung damage in mice by enhancing the activity of SOD and GSH-Px, reducing the MDA and ROS level and attenuating the upregulation of pro-inflammatory cytokine (IL-6, IL-8, IL-1ß and TNF-α), down regulating the protein expression of NF-κB, STAT3 and Caspase-3. Conclusion: Our findings suggest that the effective MECHs formula will become a novel strategy for preventing PM-induced lung injury and provide a paradigm for the development of functional foods using MECHs.

Study on polycyclic macromolecular drug solid stability: A case exploration of methylcobalamin.

As one of derivatives of Vitamin B12, methylcobalamin (MeCbl) is an indispensable "Life Element" and plays an essential role in maintaining human normal physiology function and clinical medicine application. Because of the intricate molecular structure, strong hygroscopicity and optical instability, maintaining its solid stability is a great challenge in pharmaceutical preparation. Based on the structure features of MeCbl hydrates, this study explored the drug solid stability by designing solid-solid phase transformation (SSPT) experiments. Three hydrate powders of MeCbl that had special structure with isolated site and channel water molecules were discovered. It was found that drying condition and surrounding humidity were controlling factors influencing the final solid form. The inter-conversion relations relevant to heating-induced and humidity-induced structure changes were established among the three hydrate powders. Powder X-ray diffraction, thermogravimetric analysis, differential scanning calorimetry, high performance liquid chromatography and dynamic vapor sorption were used to characterize the differences and related properties of stably prepared MeCbl hydrate powders. The particle size of product could be regulated and controlled by optimizing operating conditions of crystallization process, where ultrasound-assisted and seeding-introduced were applied as promising strategies to enhance solution crystallization process. This study opens up the possibility for the stable preparation and large-scale production of polycyclic macromolecular bulk drugs like methylcobalamin.

Influence of Different Treatments on the Structure and Conversion of Silicon Species in Rice Straw to Tetraethyl Orthosilicate (TEOS).

The production of tetraethyl orthosilicate (TEOS) from biomass provides a new way for TEOS production and biomass valorization. In this study, rice straw was treated using different fractionation methods, and the content, state, and reactivity of Si in the treated samples were investigated. It was found that acid treatment and ethanol extraction kept most Si in the biomass, while alkali treatment caused significant Si loss. Si was mainly present in the SiOx , Si-O-C, and Si-O-Si states in the surface of raw rice straw, cellulose and Klason lignin. The results showed that the Si-O-Si state in rice straw was beneficial for the formation of TEOS. The removal of lipids from rice straw facilitated the production of TEOS, giving the highest TEOS yield of 76.2 %. In contrast, the production of TEOS from other samples became difficult; the simultaneous conversion of the three organic components of rice straw also facilitated the production of TEOS.

Visual Gait Analysis Based on UE4.

With the development of artificial intelligence technology, virtual reality technology has been widely used in the medical and entertainment fields, as well as other fields. This study is supported by the 3D modeling platform in UE4 platform technology and designs a 3D pose model based on inertial sensors through blueprint language and C++ programming. It can vividly display changes in gait, as well as changes in angles and displacements of 12 parts such as the big and small legs and arms. It can be used to combine with the module of capturing motion which is based on inertial sensors to display the 3D posture of the human body in real-time and analyze the motion data. Each part of the model contains an independent coordinate system, which can analyze the angle and displacement changes of any part of the model. All joints of the model are interrelated, the motion data can be automatically calibrated and corrected, and errors measured by an inertial sensor can be compensated, so that each joint of the model will not separate from the whole model and there will not occur actions that against the human body's structures, improving the accuracy of the data. The 3D pose model designed in this study can correct motion data in real time and display the human body's motion posture, which has great application prospects in the field of gait analysis.

Online Monitoring of Seawater Carbon Dioxide Based on an Infrared Rear Beam Splitter.

The ocean is one of the most extensive ecosystems on Earth and can absorb large amounts of carbon dioxide. Changes in seawater carbon dioxide concentrations are one of the most important factors affecting marine ecosystems. Excess carbon dioxide can lead to ocean acidification, threatening the stability of marine ecosystems and species diversity. Dissolved carbon dioxide detection in seawater has great scientific significance. Conducting online monitoring of seawater carbon dioxide can help to understand the health status of marine ecosystems and to protect marine ecosystems. Current seawater detection equipment is large and costly. This study designed a low-cost infrared carbon dioxide detection system based on molecular theory. Using the HITRAN database, the absorption spectra and coefficients of carbon dioxide molecules under different conditions were calculated and derived, and a wavelength of 2361 cm-1 was selected as the measurement channel for carbon dioxide. In addition, considering the interference effect of direct light, an infrared post-splitting method was proposed to eliminate the interference of light and improve the detection accuracy of the system. The system was designed for the online monitoring of carbon dioxide in seawater, including a peristaltic pump to accelerate gas-liquid separation, an optical path structure, and carbon dioxide concentration inversion. The experimental results showed that the standard deviation of the gas test is 3.05, the standard deviation of the seawater test is 6.04, and the error range is within 20 ppm. The system can be flexibly deployed and has good stability and portability, which can meet the needs of the online monitoring of seawater carbon dioxide concentration.

Few-Shot Relation Extraction With Dual Graph Neural Network Interaction.

Recent advances in relation extraction with deep neural architectures have achieved excellent performance. However, current models still suffer from two main drawbacks: 1) they require enormous volumes of training data to avoid model overfitting and 2) there is a sharp decrease in performance when the data distribution during training and testing shift from one domain to the other. It is thus vital to reduce the data requirement in training and explicitly model the distribution difference when transferring knowledge from one domain to another. In this work, we concentrate on few-shot relation extraction under domain adaptation settings. Specifically, we propose, a novel graph neural network (GNN) based approach for few-shot relation extraction. leverages an edge-labeling dual graph (i.e. an instance graph and a distribution graph) to explicitly model the intraclass similarity and interclass dissimilarity in each individual graph, as well as the instance-level and distribution-level relations across graphs. A dual graph interaction mechanism is proposed to adequately fuse the information between the two graphs in a cyclic flow manner. We extensively evaluate on FewRel1.0 and FewRel2.0 benchmarks under four few-shot configurations. The experimental results demonstrate that can match or outperform previously published approaches. We also perform experiments to further investigate the parameter settings and architectural choices, and we offer a qualitative analysis.

Advanced synergetic nitrogen removal of municipal wastewater using oxidation products of refractory organic matters in secondary effluent by biogenic manganese oxides as carbon source.

Due to the high operational cost and secondary pollution of the conventional advanced nitrogen removal of municipal wastewater, a novel concept and technique of advanced synergetic nitrogen removal of partial-denitrification anammox and denitrification was proposed, which used the oxidation products of refractory organic matters in the secondary effluent of municipal wastewater treatment plant (MWWTP) by biogenic manganese oxides (BMOs) as carbon source. When the influent NH4+-N in the denitrifying filter was about 1.0, 2.0, 3.0, 4.0, 5.0 and 7.0 mg/L, total nitrogen (TN) in the effluent decreased from about 22 mg/L to 11.00, 7.85, 6.85, 5.20, 4.15 and 2.09 mg/L, and the corresponding removal rate was 49.15, 64.82, 69.40, 76.70, 81.36 and 90.58%, respectively. The proportional contribution of the partial-denitrification anammox pathway to the TN removal was 12.00, 26.45, 39.70, 46.04, 54.97 and 64.01%, and the actual CODcr consumption of removing 1 mg TN was 0.75, 1.43, 1.26, 1.17, 1.08 and 0.99 mg, respectively, which was much lower than the theoretical CODcr consumption of denitrification. Furthermore, CODcr in the effluent decreased to 8.12 mg/L with a removal rate of 72.40%, and the removed organic matters were mainly non-fluorescent organic matters. Kinds of denitrifying bacteria, anammox bacteria, hydrolytic bacteria and manganese oxidizing bacteria (MnOB) were identified in the denitrifying filter, which demonstrated that the advanced synergetic nitrogen removal was achieved. This novel technology presented the advantages of high efficiency of TN and CODcr removal, low operational cost and no secondary pollution.

Prototype Completion for Few-Shot Learning.

Few-shot learning (FSL) aims to recognize novel classes with few examples. Pre-training based methods effectively tackle the problem by pre-training a feature extractor and then fine-tuning it through the nearest centroid based meta-learning. However, results show that the fine-tuning step makes marginal improvements. In this paper, 1) we figure out the reason, i.e., in the pre-trained feature space, the base classes already form compact clusters while novel classes spread as groups with large variances, which implies that fine-tuning feature extractor is less meaningful; 2) instead of fine-tuning feature extractor, we focus on estimating more representative prototypes. Consequently, we propose a novel prototype completion based meta-learning framework. This framework first introduces primitive knowledge (i.e., class-level part or attribute annotations) and extracts representative features for seen attributes as priors. Second, a part/attribute transfer network is designed to learn to infer the representative features for unseen attributes as supplementary priors. Finally, a prototype completion network is devised to learn to complete prototypes with these priors. Moreover, to avoid the prototype completion error, we further develop a Gaussian based prototype fusion strategy that fuses the mean-based and completed prototypes by exploiting the unlabeled samples. At last, we also develop an economic prototype completion version for FSL, which does not need to collect primitive knowledge, for a fair comparison with existing FSL methods without external knowledge. Extensive experiments show that our method: i) obtains more accurate prototypes; ii) achieves superior performance on both inductive and transductive FSL settings.

CCP5 and CCP6 retain CP110 and negatively regulate ciliogenesis.

BACKGROUND: The axonemal microtubules of primary cilium undergo a conserved protein posttranslational modification (PTM) - polyglutamylation. This reversible procedure is processed by tubulin tyrosine ligase-like polyglutamylases to form secondary polyglutamate side chains, which are metabolized by the 6-member cytosolic carboxypeptidase (CCP) family. Although polyglutamylation modifying enzymes have been linked to ciliary architecture and motility, it was unknown whether they also play a role in ciliogenesis. RESULTS: In this study, we found that CCP5 expression is transiently downregulated upon the initiation of ciliogenesis, but recovered after cilia are formed. Overexpression of CCP5 inhibited ciliogenesis, suggesting that a transient downregulation of CCP5 expression is required for ciliation initiation. Interestingly, the inhibitory effect of CCP5 on ciliogenesis does not rely on its enzyme activity. Among other 3 CCP members tested, only CCP6 can similarly suppress ciliogenesis. Using CoIP-MS analysis, we identified a protein that potentially interacts with CCP - CP110, a known negative regulator of ciliogenesis, whose degradation at the distal end of mother centriole permits cilia assembly. We found that both CCP5 and CCP6 can modulate CP110 level. Particularly, CCP5 interacts with CP110 through its N-terminus. Loss of CCP5 or CCP6 led to the disappearance of CP110 at the mother centriole and abnormally increased ciliation in cycling RPE-1 cells. Co-depletion of CCP5 and CCP6 synergized this abnormal ciliation, suggesting their partially overlapped function in suppressing cilia formation in cycling cells. In contrast, co-depletion of the two enzymes did not further increase the length of cilia, although CCP5 and CCP6 differentially regulate polyglutamate side-chain length of ciliary axoneme and both contribute to limiting cilia length, suggesting that they may share a common pathway in cilia length control. Through inducing the overexpression of CCP5 or CCP6 at different stages of ciliogenesis, we further demonstrated that CCP5 or CCP6 inhibited cilia formation before ciliogenesis, while shortened the length of cilia after cilia formation. CONCLUSION: These findings reveal the dual role of CCP5 and CCP6. In addition to regulating cilia length, they also retain CP110 level to suppress cilia formation in cycling cells, pointing to a novel regulatory mechanism for ciliogenesis mediated by demodifying enzymes of a conserved ciliary PTM, polyglutamylation.

WDMNet: Modeling diverse variations of regional wind speed for multi-step predictions.

Regional wind speed prediction plays an important role in the development of wind power, which is usually recorded in the form of two orthogonal components, namely U-wind and V-wind. The regional wind speed has the characteristics of diverse variations, which are reflected in three aspects: (1) The spatially diverse variations of regional wind speed indicate that wind speed has different dynamic patterns at different positions; (2) The distinct variations between U-wind and V-wind denote that U-wind and V-wind at the same position exhibit different dynamic patterns; (3) The non-stationary variations of wind speed represent that the intermittent and chaotic nature of wind speed. In this paper, we propose a novel framework named Wind Dynamics Modeling Network (WDMNet) to model the diverse variations of regional wind speed and make accurate multi-step predictions. To jointly capture the spatially diverse variations and the distinct variations between U-wind and V-wind, WDMNet leverages a new neural block called Involution Gated Recurrent Unit Partial Differential Equation (Inv-GRU-PDE) as its key component. The block adopts involution to model spatially diverse variations and separately constructs hidden driven PDEs of U-wind and V-wind. The construction of PDEs in this block is achieved by a new Involution PDE (InvPDE) layers. Besides, a deep data-driven model is also introduced in Inv-GRU-PDE block as the complement to the constructed hidden PDEs for sufficiently modeling regional wind dynamics. Finally, to effectively capture the non-stationary variations of wind speed, WDMNet follows a time-variant structure for multi-step predictions. Comprehensive experiments have been conducted on two real-world datasets. Experimental results demonstrate the effectiveness and superiority of the proposed method over state-of-the-art techniques.

Span-based single-stage joint entity-relation extraction model.

Extracting entities and relations from the unstructured text has attracted increasing attention in recent years. The existing work has achieved considerable results, yet it is difficult to solve entity overlap and exposure bias. To address cascading errors, exposure bias, and entity overlap in existing entity relation extraction approaches, we propose a joint entity relation extraction model (SMHS) based on a span-level multi-head selection mechanism, transforming entity relation extraction into a span-level multi-head selection problem. Our model uses span-tagger and span-embedding to construct span semantic vectors, utilizes LSTM and multi-head self-attention mechanism for span feature extraction, multi-head selection mechanism for span-level relation decoding, and introduces span classification task for multi-task learning to decode out the relation triad in a single-stage. Experiments on the classic English dataset NYT and the publicly available Chinese relationship extraction dataset DuIE 2.0 show that this method achieves better results than the baseline method, which verifies the effectiveness of this method. Source code and data are published here(https://github.com/Beno-waxgourd/NLP.git).