Search

In network analysis, methods for identifying a community structure of ten involve optimizing a specific objective function to achieve a single optimal allocation from network nodes to communities. In practice, however, we often encounter multiple division schemes with high-quality scores close to the overall optimum. An accurate depiction of the community structure is more appropriately achieved by a series of high-quality division schemes rather than relying solely on a single optimal solution. However, such a collection of network divisions may be challenging to interpret, as its size may rapidly expand to hundreds or even thousands. To this end, we propose a representative community detection algorithm for attribute networks. By clustering similar network partitions and selecting representative partitions from each cluster, we can comprehensively reveal the diversity of network community structures and provide partition results with a more global perspective. Network partitioning experiments on natural and artificial datasets demonstrate that our proposed method performs better than advanced methods.

The unsupervised attribute graph embedding technique aims to learn low-dimensional node embedding using neighborhood topology and attribute information under unlabeled data. Current unsupervised models are mostly based on graph self-encoders, but full-batch training limits the scalability of the model and ignores attribute integrity when reconstructing the topology. In order to solve the above problems while considering the unsupervised learning of the model and full use of node information, this paper proposes a graph neural network architecture based on a graph self-encoder to capture the nonlinearity of the attribute graph data, and an attribute graph embedding algorithm that explicitly models the influence of neighborhood information using a multi-level attention mechanism. Specifically, the proposed algorithm fuses topology information and attribute information using a lightweight sampling strategy, constructs an unbiased graph self-encoder on the sampled graph, implements topology aggregation and attribute aggregation, respectively, models the correlation between topology embedding and attribute embedding, and considers multi-level loss terms.

We are delighted to dedicate this Special Issue to a leading figure in actinoid chemistry, Prof [...]

Abstract This paper presents an interesting case study on Legacy Data Integration (LDI for short) for a Regional Cloud Arbitration Court. Due to the inconsistent structure and presentation, legacy arbitration cases can hardly integrate into the Cloud Court unless processed manually. In this study, we propose an AI-enabled LDI method to replace the costly manual approach and ensure privacy protection during the process. We trained AI models to replace tasks such as reading and understanding legacy cases, removing privacy information, composing new case records, and inputting them through the system interfaces. Our approach employs Optical Character Recognition (OCR), text classification, and Named Entity Recognition (NER) to transform legacy data into a system format. We applied our method to a Cloud Arbitration Court in Liaoning Province, China, and achieved a comparable privacy filtering effect while retaining the maximum amount of information. Our method demonstrated similar effectiveness as the manual LDI, but with greater efficiency, saving 90% of the workforce and achieving a 60%-70% information extraction rate compared to manual work. With the increasing development of informationalization and intelligentization in judgment and arbitration, many courts are adopting ABC technologies, namely Artificial intelligence, Big data, and Cloud computing, to build the court system. Our method provides a practical reference for integrating legal data into the system.

The recent few decades witnessed a quick growth in our knowledge in actinoid chemistry, particularly in actinoids’ behaviors in catalysis and biologic systems [...]

Nowadays, the distributed tree-based index has been widely adopted to process queries on large-scale data. Index placement is an essential part of distributed tree-based index management. Current research on index placement relies on predefined rules, such as user-supplied functions, load balance, and data location. They ignore the distance between the index position and the query data (denoted as locality), which significantly determines the index placement efficiency. However, the locality relies on several factors related to unpredictable queries, and it is difficult to consider all of the factors. Measuring query performance can comprehensively consider the above query-related factors. Therefore, we propose a novel method that uses deep reinforcement learning (DRL) to solve the problem and adopts the query performance as evaluation feedback in DRL. Based on this idea, we design a tool called LiPTool that learns the optimal index placement. Specifically, LiPTool employs deep reinforcement learning (DRL) to find the response server for every query, maximizing the average query performance of the workload. Then, LiPTool builds the index on each server according to the corresponding queries. Finally, LiPTool gives the suggestion of index placement. Moreover, this paper compares LiPTool with the other two methods on multiple datasets, and the results show that LiPTool doubles the query performance. We believe that LiPTool is a good solution for dynamically adjusting index placement in autonomous database management.

Oxidative dehydrogenation of alkanes over boron nitride catalysts provides a new opportunity for efficient olefin production. Here, the authors discover in situ formed olefins can promote parent alkane conversion, and achieve activation of ethane by propane-derived olefins at a lower temperature.

Self-supervised learning is a new machine learning method that does not rely on manually labeled data, and learns from rich unlabeled data itself by designing agent tasks using the input data as supervision to obtain a more generalized representation for application in downstream tasks. However, the current self-supervised learning suffers from the problem of relying on the selection and number of negative samples and the problem of sample bias phenomenon after graph data augmentation. In this paper, we investigate the above problems and propose a corresponding solution, proposing a graph contrastive learning algorithm without negative samples. The model uses matrix sketching in the implicit space for feature augmentation to reduce sample bias and iteratively trains the mutual correlation matrix of two viewpoints by drawing closer to the distance of the constant matrix as the objective function. This method does not require techniques such as negative samples, gradient stopping, and momentum updating to prevent self-supervised model collapse. This method is compared with 10 graph representation learning algorithms on four datasets for node classification tasks, and the experimental results show that the algorithm proposed in this paper achieves good results.

In real-world scenarios, obtaining the relationships between nodes is often challenging, resulting in incomplete network topology. This limitation significantly reduces the applicability of community detection methods, particularly neighborhood aggregation-based approaches, on structurally incomplete networks. Therefore, in this situation, it is crucial to obtain meaningful community information from the limited network structure. To address this challenge, the LPGSE algorithm was designed and implemented, which includes four parts: link prediction, structure observation, network estimation, and community partitioning. LPGSE demonstrated its performance in community detection in structurally incomplete networks with 10% missing edges on multiple datasets. Compared with traditional community detection algorithms, LPGSE achieved improvements in NMI and ARI metrics of 1.5781% to 29.0780% and 0.4332% to 31.9820%, respectively. Compared with similar community detection algorithms for structurally incomplete networks, LPGSE also outperformed other algorithms on all datasets. In addition, different edge-missing ratio settings were also attempted, and the performance of different algorithms in these situations was compared and analyzed. The results showed that the algorithm can still maintain high accuracy and stability in community detection across different edge-missing ratios.

The North China Plain plays a pivotal role in China’s crop production, contributing to 30% of the maize yield. Nevertheless, summer maize in this region faces challenges due to climatic constraints characterized by concurrent high temperatures and rainfall during the growing season, resulting in a relatively high evapotranspiration rate. In this study, we explored eight soil moisture-based threshold irrigation strategies, consisting of two upper limits and four lower limits, along with a rainfed mode (E). The upper and lower irrigation limits are expressed as a percentage of the field’s water-holding capacity (sfc). For the four full irrigation modes (A1, A2, A3, A4), the lower limits were set at 0.6 sfc, 0.6 sfc, 0.5 sfc, and 0.5 sfc, respectively. The upper limits were defined at two levels: 0.8 sfc for A1 and A2 and sfc for A3 and A4. Similarly, for the four deficit irrigation modes (B1, B2, B3, B4), the lower limits were established at 0.4 sfc, 0.4 sfc, 0.3 sfc, and 0.3 sfc, respectively, with the upper limits set at two levels: 0.8 sfc for B1 and B2 and the full sfc for B3 and B4. To investigate the impact of rainfall and potential evapotranspiration on these irrigation modes under long-term fluctuations, we employed a stochastic framework that probabilistically linked rainfall events and irrigation applications. The Monte Carlo method was employed to simulate a long-term series (4000a) of rainfall parameters and evapotranspiration using 62 years of meteorological data from the Xinxiang region, situated in the southern part of the North China Plain. Results showed that the relative yield and net irrigation water requirement of summer maize decreased with decreasing irrigation lower limits. Additionally, the interannual variation of rainfall parameters and evapotranspiration during the growing season were remarkable, which led to the lowest relative yield of the rainfed mode (E) aligned with a larger interannual difference. According to the simulation results, mode A4 (irrigation lower limit equals 0.5 sfc, irrigation upper limit equals 0.8 sfc) could be adopted for adequate water resources. Conversely, mode B2 is more suitable for a lack of water resources.

The separation of trivalent actinides and lanthanides is a key step in the sustainable development of nuclear energy, and it is currently mainly realized via liquid–liquid extraction techniques. The underlying mechanism is complicated and remains ambiguous, which hinders the further development of extraction. Herein, to better understand the mechanism of the extraction, the contributing factors for the extraction are discussed (specifically, the sulfur-donating ligand, Cyanex301) by combing molecular dynamics simulations and experiments. This work is expected to contribute to improve our systematic understanding on a molecular scale of the extraction of lanthanides and actinides, and to assist in the extensive studies on the design and optimization of novel ligands with improved performance.

Direct ascorbic acid fuel cells (DAAFCs) employ biocompatible ascorbic acid (AA) as fuel, allowing convenient storage, transportation, and fueling as well as avoiding fuel crossover. The AA oxidation reaction (AAOR) largely governs the performance of DAAFCs. However, AAOR electrocatalysts currently have low activity, and state-of-the-art ones are limited to carbon black. Herein, we report the synthesis of an unprecedented AAOR electrocatalyst comprising 3.9 ± 1.1 nm CeO2 nanoparticles evenly distributed on carbon black simply by the wet chemical precipitation of Ce(OH)3 and a subsequent heat treatment. The resultant CeO2/C shows a remarkable AAOR activity with a peak current density of 13.1 mA cm−2, which is 1.7 times of that of carbon black (7.67 mA cm−2). According to X-ray photoelectron spectroscopy (XPS), the surface Ce3+ of CeO2 appears to contribute to the AAOR activity. Furthermore, our density functional theory (DFT) calculation reveals that that the proton of the hydroxyl group of AA can easily migrate to the bridging O sites of CeO2, resulting in a faster AAOR with respect to the pristine carbon, -COOH, and -C=O sites of carbon. After an i-t test, CeO2/C loses 17.8% of its initial current density, which is much superior to that of carbon black. CeO2 can capture the electrons generated by the AAOR to protect the -COOH and -C=O sites from being reduced. Finally, DAAFCs fabricated with CeO2/C exhibit a remarkable power density of 41.3 mW cm−2, which is the highest among proton-exchange-membrane-based DAAFCs in the literature.

Abstract Objectives The aim of this study was to evaluate the role of antiviral drugs in reducing the risk of developing severe illness in patients with moderate COVID-19 pneumonia. Methods This retrospective cohort study included 403 adult patients with moderate COVID-19 pneumonia who were admitted to Shenzhen Third People’s Hospital, China. The antiviral drugs arbidol, interferon alpha-1b, lopinavir–ritonavir and ribavirin were distributed to the patients for treatment. The primary endpoint of this study was the time to develop severe illness. Results Of the 462 patients admitted, 403 had moderate COVID-19 symptoms at hospital admission and were included in this study. 90 of the 403 (22.3%) patients progressed to severe illness. The use of arbidol was associated with a lower severity rate 3.5% compared to control group 30.5%, p-value

In this paper, we propose an outlier-detection algorithm for detecting network traffic anomalies based on a clustering algorithm and an autoencoder model. The BIRCH clustering algorithm is employed as the pre-algorithm of the autoencoder to pre-classify datasets with complex data distribution characteristics, while the autoencoder model is used to detect outliers based on a threshold. The proposed BIRCH-Autoencoder (BAE) algorithm has been tested on four network security datasets, KDDCUP99, UNSW-NB15, CICIDS2017, and NSL-KDD, and compared with representative algorithms. The BAE algorithm achieved average F-scores of 96.160, 81.132, and 91.424 on the KDDCUP99, UNSW-NB15, and CICIDS2017 datasets, respectively. These experimental results demonstrate that the proposed approach can effectively and accurately detect anomalous data.

Abstract The assessment of protein–ligand interactions is critical at early stage of drug discovery. Computational approaches for efficiently predicting such interactions facilitate drug development. Recently, methods based on deep learning, including structure- and sequence-based models, have achieved impressive performance on several different datasets. However, their application still suffers from a generalizability issue because of insufficient data, especially for structure based models, as well as a heterogeneity problem because of different label measurements and varying proteins across datasets. Here, we present an interpretable multi-task model to evaluate protein–ligand interaction (Multi-PLI). The model can run classification (binding or not) and regression (binding affinity) tasks concurrently by unifying different datasets. The model outperforms traditional docking and machine learning on both binary classification and regression tasks and achieves competitive results compared with some structure-based deep learning methods, even with the same training set size. Furthermore, combined with the proposed occlusion algorithm, the model can predict the important amino acids of proteins that are crucial for binding, thus providing a biological interpretation.

BackgroundPercutaneous coronary intervention (PCI) is the preferred treatment method for coronary artery diseases (CAD). This study aimed to evaluate the effectiveness and complications of the Guidezilla™ guide extension catheter I (GGEC I) in transradial coronary intervention (TRI).MethodsThis case series study included patients with CAD who underwent TRI using the GGEC I between August 2016 and January 2019 at the First Affiliated Hospital of Xi’an Jiaotong University.ResultsA total of 221 patients aged 65.1 ± 9.26 years were included. Coronary angiography results indicated that most patients (77.8%) had triple-vessel lesions, including 47.5% with chronic total occlusion (CTO). A total of 237 target lesions were treated, most being type C lesions (95.8%). The most common indication for GGEC I use was heavy calcification (67%), followed by extreme tortuosity (12.2%), extreme tortuosity and heavy calcification (10.9%), distally located lesion (4.5%), picking up the retrograde wire (3.2%), anomalous vessel origin (1.8%), and releasing the burr incarceration (0.4%). The mean operation time was 58 min, and the overall success rate was 94.1%. Four patients received a drug-coated balloon. No significant differences were found in operation time and success rate among the low (32) CAD groups based on SYNTAX score stratification (P > 0.05). Two subacute thrombosis cases each were reported perioperatively, during hospitalization, and at the 1-month follow-up.ConclusionThe GGEC I might have advantages for TRI and is unaffected by SYNTAX score stratification.

Electrochemical impedance spectroscopy (EIS) is an efficient and non-destructive test for analyzing the bioelectrochemical processes of microbial fuel cells (MFCs). The key factors limiting the output performance of an MFC can be identified by quantifying the contribution of its various internal parts to the total impedance. However, little attention has been paid to the measurement conditions and diagrammatic processes of the EIS for MFC. This review, starting with the analysis of admittance of bioelectrode, introduces conditions for the EIS measurement and summarizes the representative equivalent circuit plots for MFC. Despite the impedance from electron transfer and diffusion process, the effect of unnoticeable capacitance obtained from the Nyquist plot on MFCs performance is evaluated. Furthermore, given that distribution of relaxation times (DRT) is an emerging method for deconvoluting EIS data in the field of fuel cell, the application of DRT-analysis to MFC is reviewed here to get insight into bioelectrode reactions and monitor the biofilm formation. Generally, EIS measurement is expected to optimize the construction and compositions of MFCs to overcome the low power generation.