Network Datasets Research Articles

A Dual-Stream Cross AGFormer-GPT Network for Traffic Flow Prediction Based on Large-Scale Road Sensor Data.

Traffic flow prediction can provide important reference data for managers to maintain traffic order, and can also be based on personal travel plans for optimal route selection. On account of the development of sensors and data collection technology, large-scale road network historical data can be effectively used, but their high non-linearity makes it meaningful to establish effective prediction models. In this regard, this paper proposes a dual-stream cross AGFormer-GPT network with prompt engineering for traffic flow prediction, which integrates traffic occupancy and speed as two prompts into traffic flow in the form of cross-attention, and uniquely mines spatial correlation and temporal correlation information through the dual-stream cross structure, effectively combining the advantages of the adaptive graph neural network and large language model to improve prediction accuracy. The experimental results on two PeMS road network data sets have verified that the model has improved by about 1.2% in traffic prediction accuracy under different road networks.

The performance of the practices associated with the occurrence of severe intraventricular hemorrhage in the very premature infants: data analysis from the Chinese neonatal network

BackgroundThe occurrence of severe intraventricular hemorrhage (sIVH) was high in the very preterm infants (VPIs) in China. The management strategies significantly contributed to the occurrence of sIVH in VPIs. However, the status of the perinatal strategies associated with sIVH for VPIs was rarely described across the multiple neonatal intensive care units (NICUs) in China. We aim to investigate the characteristics of the perinatal strategies associated with sIVH for VPIs across the multiple NICUs in China.MethodsThis was a retrospective analysis of data from a prospective cohort of Chinese Neonatal Network (CHNN) dataset, enrolling infants born at 24+0—31+6 from 2019 to 2021. Eleven perinatal practices performed within the first 3 days of life were investigated including antenatal corticosteroids use, antenatal magnesium sulphate therapy, intubation at birth, placental transfusion, need for advanced resuscitation, initial inhaled gas of 100% FiO2 in delivery room, initial invasive respiratory support, surfactant and caffeine administration, early enteral feeding, and inotropes use. The performances of these practices across the multiple NICUs were investigated using the standard deviations of differences between expected probabilities and observations. The occurrence of sIVH were compared among the NICUs.ResultsA total of 24,226 infants from 55 NICUs with a mean (SD) gestational age of 29.5 (1.76) and mean (SD) birthweight of 1.31(0.32) were included. sIVH was detected in 5.1% of VPIs. The rate of the antenatal corticosteroids, MgSO4 therapy, and caffeine was 80.0%, 56.4%, and 31.5%, respectively. We observed significant relationships between sIVH and intubation at birth (AOR 1.52, 95% CI 1.13 to 1.75) and initial invasive respiratory support (AOR 2.47, 95% CI 2.15 to 2.83). The lower occurrence of sIVH (4.8%) was observed corresponding with the highest utility of standard antenatal care, the lowest utility of invasive practices, and early enteral feeding administration.ConclusionsThe current evidence-based practices were not performed in each VPI as expected among the studied Chinese NICUs. The higher utility of the invasive practices could be related to the occurrence of sIVH.

An integrated graph data privacy attack framework based on graph neural networks in IoT

SummaryKnowledge graphs contain a large amount of entity and relational data, and graph neural networks, as a class of efficient graph representation techniques based on deep learning, excel in knowledge graph modeling. However, previous neural network architectures for the most part only learn node representations and do not fully consider the heterogeneity of data. In this article, we innovatively propose a privacy attack framework based on IoT, PAFI, which is able to classify entities and relations, learn embedding representations in multi‐relational graphs, and can be applied to some existing neural network algorithms. Based on this, a fine‐grained privacy attack model, FPM, is proposed, which can perform attack operations on multiple targets, achieve selectivity of target tasks, and greatly improve the generalization ability of the attack model. In this article, the effectiveness of PAFI and FPM is demonstrated by real network datasets, and compared with previous attack methods, both of which achieve good results.

Deep learning-based 3D reconstruction from multiple images: A survey

Reconstructing the three-dimensional structure of a scene is a classic and fundamental problem in computer vision, but it has been revolutionized by recent advancements in deep machine learning. In this paper, we survey this rich and growing area. We divide the work into four main threads: 3D reconstruction from two calibrated images from a binocular camera; 3D reconstruction from more than two images taken by the same camera or more than two calibrated cameras; object-focused 3D reconstruction with relaxed camera calibration; and SLAM-based techniques. We summarize each approach along four salient dimensions: algorithmic and deep network characteristics, output representation, datasets, and quantitative comparisons among different methods. We also discuss key challenges and future directions.

Facial micro-expression recognition using stochastic graph convolutional network and dual transferred learning

Micro-expression recognition (MER) has drawn increasing attention due to its wide application in lie detection, criminal detection and psychological consultation. However, the best recognition accuracy on recent public dataset is still low compared to the accuracy of macro-expression recognition. In this paper, we propose a novel graph convolution network (GCN) for MER achieving state-of-the-art accuracy. Different to existing GCN with fixed graph structure, we define a stochastic graph structure in which some neighbors are selected randomly. As shown by numerical examples, randomness enables better feature characterization while reducing computational complexity. The whole network consists of two branches, one is the spatial branch taking micro-expression images as input, the other is the temporal branch taking optical flow images as input. Because the micro-expression dataset does not have enough images for training the GCN, we employ the transfer learning mechanism. That is, different stochastic GCNs (SGCN) have been trained by the macro-expression dataset in the source network. Then the well-trained SGCNs are transferred to the target network. It is shown that our proposed method achieves the state-of-art performance on all four well-known datasets. This paper explores stochastic GCN and transfer learning with this random structure in the MER task, which is of great importance to improve the recognition performance.

Multi-institutional validation survey on Belong.life's conversational artificial intelligence (AI) oncology mentor, "Dave.

e13596 Background: Belong.life, a global oncology social network for patients and caregivers, recently launched “Dave” the first conversational AI oncology mentor and companion. “Dave’s” objectives are to provide uninterrupted support, clarify relevant clinical issues and guide patients and caregivers with relevant and empathetic information and education in all aspects of cancer, from diagnosis to treatment protocols and side effects management. “Dave” underwent training on Belong’s unique and large datasets of patients to physicians, and patients to patients’ interactions, as well as incorporating high quality information from the latest international cancer guidelines, providing it with a robust up-to-date supportive data and a wide understanding of the patients’ cancer journey. Methods: “Dave’s” responses to inquiries from Belong members were subjected to a validation survey conducted by eight oncologists, each specializing in various solid and haematological cancers and affiliated with several medical institutions. From the social network datasets, 471 questions from patients and caregivers were randomly selected and categorized into groups, including Breast, Gastrointestinal and Pancreatic, Genito-urinary, Bone, Haematological cancers, and Radiation therapy. The oncologists assessed the AI mentor's replies for their relevance and helpfulness, aligning them with evidence-based medicine, current recommendations, and guidelines. Results: Results were categorized into positive or negative assessments. Impressively, 432 out of 471 “Dave’s” responses (91.8%) received positive validation for providing suitable and relevant recommendations. Conversely, 39/471 responses (8.2%) received a negative validation, of which only 5/471 (1%) were graded as not at all helpful or relevant. Minor variations were observed within diagnostic groups, with positive validation rates as follows: Genito-urinary 100%, Breast 98.4%, Musculoskeletal 97%, Radiation therapy 88%, Gastrointestinal and Pancreatic 86%, and Haematological cancers 84%. Conclusions: The documented results on 471 patients’ posts are very promising, indicating that “Dave” provides valuable and reliable recommendations, achieving a positive validation rate of nearly 92%, with only 1% graded as not at all helpful or relevant. The findings also offer insights into the AI mentor's performance across distinct medical domains. This validation study provides a solid foundation and adds confirmation that the addition of an AI oncology mentor and companion, like “Dave”, improves patients’ knowledge and coping mechanisms and provides helpful and relevant guidance during their cancer journey, while supporting physicians in the daily management of their cancer patients.

A cascaded GRU-based stereoscopic matching network for precise plank measurement

Wooden plank images in industrial measurements often contain numerous textureless areas. Furthermore, due to the thin plate structure, the three-dimensional (3D) disparity of these planks is predominantly confined to a narrow range. Consequently, achieving accurate 3D matching of wooden plank images has consistently presented a challenging task within the industry. In recent years, deep learning has progressively supplanted traditional stereo matching methods due to its inherent advantages, including rapid inference and end-to-end processing. Nonetheless, the acquisition of datasets for stereo matching networks poses an additional challenge, primarily attributable to the difficulty in obtaining accurate disparity data. Thus, this paper presents a novel stereo matching method incorporating three key innovations. Firstly, an enhanced gated recurrent unit network is introduced, accompanied by a redesigned structure to achieve higher matching accuracy. Secondly, an efficient preprocessing module is proposed, aimed at improving the algorithm’s efficiency. Lastly, in response to the challenges posed by datasets acquisition, we innovatively employed image simulation software to obtain a high-quality simulated dataset of wooden planks. To assess the feasibility of our approach, we conducted both simulated and real experiments. The experiments results clearly exhibit the superiority of our method when compared to existing approaches in terms of both stability and accuracy. In the simulation experiment, our method attained a bad1.0 score of 2.1% (compared to the baseline method’s 9.76%); In the real experiment, our method achieved an average error of 0.104 mm (compared to the baseline method’s 0.268 mm). It is worth noting that our study aims to address the challenge of acquiring datasets for deep learning and bridging the gap between simulated and real data, resulting in increased applicability of deep learning in more industrial measurement domains.

Space-time image velocimetry in blurred scenes based on BSTI-DCGAN data augmentation

Due to the limited sample quantity and the complex data collection process of the blurred space-time image (BSTI) dataset, the deep learning-based space-time image velocimetry (STIV) results in larger errors when applied to blurry videos. To enhance the measurement accuracy, we propose the use of STIV in blurred scenes based on BSTI-deep convolutional generative adversarial network (DCGAN) data augmentation. Firstly, BSTI-DCGAN is developed based on the DCGAN. This network utilizes a bilinear interpolation-convolution module for upsampling and integrates coordinated attention and multi-concatenation attention to enhance the resemblance between generated and real images. Next, further expanding the dataset by using artificially synthesized space-time images subsequently, all space-time images are transformed into spectrograms to create a training dataset for the classification network. Finally, the primary spectral direction is detected using the classification network. The experimental results indicate that our approach effectively augments the dataset and improves the accuracy of practical measurements. Under the condition of video blur, the relative errors of the average flow velocity and discharge are 3.92% and 2.72%, respectively.

Key node identification for a network topology using hierarchical comprehensive importance coefficients.

Key nodes are similar to important hubs in a network structure, which can directly determine the robustness and stability of the network. By effectively identifying and protecting these critical nodes, the robustness of the network can be improved, making it more resistant to external interference and attacks. There are various topology analysis methods for a given network, but key node identification methods often focus on either local attributes or global attributes. Designing an algorithm that combines both attributes can improve the accuracy of key node identification. In this paper, the constraint coefficient of a weakly connected network is calculated based on the Salton indicator, and a hierarchical tenacity global coefficient is obtained by an improved K-Shell decomposition method. Then, a hierarchical comprehensive key node identification algorithm is proposed which can comprehensively indicate the local and global attributes of the network nodes. Experimental results on real network datasets show that the proposed algorithm outperforms the other classic algorithms in terms of connectivity, average remaining edges, sensitivity and monotonicity.

Rapid assessment of cosmic radiation exposure in aviation based on BP neural network method.

Cosmic radiation exposure is one of the important health concerns for aircrews. In this work, we constructed a back propagation neural network model for the real-time and rapid assessment of cosmic radiation exposure to the public in aviation. The multi-dimensional dataset for this neural network was created from modeling the process of cosmic ray transportation in magnetic field by geomagnetic cutoff rigidity method and air shower simulation by a Monte Carlo based Geant4 code. The dataset was characterized by parameters including cosmic ray energy spectrum, Kp-index, coordinated universal time, altitude, latitude, and longitude. The effective dose and dose rate was finally converted from the particle fluxes at flight position by the neural network. This work shows a good agreement with other models from International Civil Aviation Organization. It is also illustrated that the effective dose rate by galactic cosmic ray is <10μSvh-1 and the value during ground level enhancement (GLE) 42 is 4~10 times larger on the routes calculated in this work. In GLE 69, the effective dose rate reaches several mSvh-1 in the polar region. Based on this model, a real-time warning system is achieved.

Spectral co-clustering in multi-layer directed networks

Modern network analysis often involves multi-layer network data in which the nodes are aligned, and the edges on each layer represent one of the multiple relations among the nodes. Current literature on multi-layer network data is mostly limited to undirected relations. However, direct relations are more common and may introduce extra information. This study focuses on community detection (or clustering) in multi-layer directed networks. To take into account the asymmetry, a novel spectral-co-clustering-based algorithm is developed to detect co-clusters, which capture the sending patterns and receiving patterns of nodes, respectively. Specifically, the eigendecomposition of the debiased sum of Gram matrices over the layer-wise adjacency matrices is computed, followed by the k-means, where the sum of Gram matrices is used to avoid possible cancellation of clusters caused by direct summation. Theoretical analysis of the algorithm under the multi-layer stochastic co-block model is provided, where the common assumption that the cluster number is coupled with the rank of the model is relaxed. After a systematic analysis of the eigenvectors of the population version algorithm, the misclassification rates are derived, which show that multi-layers would bring benefits to the clustering performance. The experimental results of simulated data corroborate the theoretical predictions, and the analysis of a real-world trade network dataset provides interpretable results.

CYP17A1 Network Analysis in Ovarian Serous Cystadenocarcinoma for Retrieval of Polycystic ovaries Targets

Background/Aims: CYP17A1 is great metabolic switch for androgen overproduction which is hallmark of polycystic ovary syndrome (PCOS)initiation and progression. There is an urgent need to determine CYP17A1 mediated set of metabolic therapeutic targets for PCOS to control androgen synthesis with wide range of molecular options. Methodology: We apply rational in silico approach for determination of PCOS comprehensive set of drug targets. First, we retrieve CYP17A1 network dataset from STRING database (https://string-db.org/) by querying CYP17A1 name that gives us updated 30 nodes containing network with unique options of enrichment analysis and module extraction. The enrichment analysis determines CYP17A1 network involvement in steroidogenesis process with carcinogenesis and drug metabolism. We select ovarian serous cystadenocarcinoma dataset from cBioPortal server (https://www.cbioportal.org/) for CYP17A1 network differential analysis. Results: In this study, several steroid synthesis pathway members showed overexpression including SRD5A1, AKR1C3, CYP11B1, CYP11B2, CYP7A1, AKR1C1, AKR1D1, CYP7B1, CYP21A2, POR and HSD17B8 and are ideal biomarkers that provide cell cycle energy requirements for ovarian carcinoma. Few anti-androgenic members such as HSD17B2, STS, SULT2B1 and CYB5A showed down regulation that predicts the impact of hyper androgenemia on carcinogenesis. Drug metabolism components also showed up regulation which can be potential biomarkers for drug resistance in chemotherapies. Conclusion: Our work suggests androgen and its synthesis pathway paramount in tumorigenesis and is an excellent therapeutic target in ovarian carcinoma. In future, validation of CYP17A1 network as a signature in both ovarian serous cystadenocarcinoma and PCOS dataset may lead to novel shared therapeutic combinations and tremendous syndrome-syndrome molecular linkage for personalized medicine.

Network Fairness Ambivalence: When does social network capital mitigate or amplify unfairness?

Social networks inherit societal biases present across lines of gender, race, socioeconomic status, and other factors. Networks can structurally perpetuate unequal access to information and opportunities through homophilous dynamics. While there is substantial knowledge about inequity in the diffusion of opportunities in a network where nodes seek them from their immediate neighbors, much less is known when considering beyond that first hop. In this paper, we leverage recent mathematical analysis of network fairness to prove that enabling simple multi-hop dissemination can reduce inequity towards a minority group in the network as long as homophily is sufficiently weak. Otherwise, our necessary and sufficient condition proves that multi-hop dissemination strategies amplify the bias already found amongst considering direct neighbors. We empirically validate these results on four social network datasets as well as present an example of a key application of our findings with a scenario of individuals who leverage their personal network to seek job referrals. Our results suggest that online platforms designing algorithms to promote opportunities to multi-hop connections must carefully take into account network metrics measuring group size and homophily in order to avoid amplifying bias against marginalized groups on their platforms.

Asymmetric influence-based superposed random walk link prediction algorithm in complex networks

Random walk-based link prediction algorithms have achieved desirable results for complex network mining, but in these algorithms, the transition probability of particles usually only considers node degrees, resulting in particles being able to randomly select adjacent nodes for random walks in an equal probability manner, to solve this problem, the asymmetric influence-based superposed random walk link prediction algorithm is proposed in this paper. This algorithm encourages particles to choose the next node at each step of the random walk process based on the asymmetric influence between nodes. To this end, we fully consider the topological information around each node and propose the asymmetric influence between nodes. Then, an adjustable parameter is applied to normalize the degree of nodes and the asymmetric influence between nodes into transition probability. Based on this, the proposed new transition probability is applied to superposed random walk process to measure the similarity between all nodes in the network. Empirical experiments are conducted on 16 real-world network datasets such as social network, ecology network, and animal network. The experimental results show that the proposed algorithm has high prediction accuracy in most network, compared with 10 benchmark indices.

Module-based graph pooling for graph classification

Graph Neural Network (GNN) models are recently proposed to process the graph-structured data for the learning tasks on graphs, e.g., node classification, link prediction, and so on. This work focuses on the graph classification task, aiming to obtain the graph representation and predict the class label for a graph. Existing works proposed applying graph pooling to obtain graph embedding but still suffer from several issues. First, node embeddings are generated according to the topological information of the whole graph, but ignoring the local isomorphic substructures commonly seen in bioinformatics and chemistry. Another limitation arises when aggregating node embeddings. The hard assignment obtained through clustering algorithms, which rely on preset and fixed parameters instead of considering the graph’s properties adaptively, restricts the flexibility in handling graphs of varying scales. To address the above problems, a module-based graph pooling framework (MGPool) is proposed in this work. Inspired by the rules of bioinformatics, MGPool assumes that a graph consists of multiple modules (also known as sub structures), which are identified based on the natural organization of the graph rather than the hard allocation of nodes. Benefiting from the hypothesis, MGPool generates node embeddings from graph-view and module-view, which is capable to capture global graph information and local isomorphic information respectively. Then module-level pooling is used to capture the intra-module information, while the inter-module information in terms of the correlation between modules is obtained through graph-level pooling. Finally, an entropy-based weighting mechanism is proposed to adjust the modules’ weights for the graph aggregation. Experiments conducted on bioinformatics benchmark datasets demonstrate the effectiveness of MGPool by outperforming other state-of-the-art graph pooling methods. For social network datasets, MGPool also provides competitive performance. Moreover, the visualization of module entropy weights is given to reveal the interpretability of the model.

Node classification in complex networks based on multi-view debiased contrastive learning

In complex networks, contrastive learning has emerged as a crucial technique for acquiring discriminative representations from graph data. Maximizing the similarity among relevant sample pairs while minimizing that among irrelevant pairs is pivotal in contrastive learning. Therefore, careful consideration must be given to the design of sample pairs in contrastive learning. However, existing node-level self-supervised contrastive learning often treats the enhanced representation of a central node as a positive sample, while considering representations of all other nodes as negative samples. This approach can lead to conflicts in downstream tasks on some graph data, as nodes of the same class are treated as negatives during learning. Precision in sample pair design is essential for enhancing the performance of contrastive learning. To address this issue, this paper introduces a negative sample debiased sampling contrastive learning (NDSCL), specifically tailored for node classification tasks. In particular, this method integrates contrastive learning with semi-supervised learning. A trained classifier assigns pseudo-labels to unlabeled data, and debiased sampling is applied to negative samples. Unlike other methods that focus on negative sample selection, NDSCL also addresses the imbalance in pseudo-label distribution by employing debiasing techniques. Finally, in conjunction with diffusion augmentation, the model is provided with diverse views as inputs to maximize the retention of underlying semantic information. Experimental results demonstrate that the proposed model significantly outperforms baseline models in node-level classification tasks across multiple network datasets. Moreover, the model not only enhances accuracy but also improves computational speed and memory requirements for handling large-scale graph data structures.

Predicting severe intraventricular hemorrhage or early death using machine learning algorithms in VLBWI of the Korean Neonatal Network Database

Severe intraventricular hemorrhage (IVH) in premature infants can lead to serious neurological complications. This retrospective cohort study used the Korean Neonatal Network (KNN) dataset to develop prediction models for severe IVH or early death in very-low-birth-weight infants (VLBWIs) using machine-learning algorithms. The study included VLBWIs registered in the KNN database. The outcome was the diagnosis of IVH Grades 3–4 or death within one week of birth. Predictors were categorized into three groups based on their observed stage during the perinatal period. The dataset was divided into derivation and validation sets at an 8:2 ratio. Models were built using Logistic Regression with Ridge Regulation (LR), Random Forest, and eXtreme Gradient Boosting (XGB). Stage 1 models, based on predictors observed before birth, exhibited similar performance. Stage 2 models, based on predictors observed up to one hour after birth, showed improved performance in all models compared to Stage 1 models. Stage 3 models, based on predictors observed up to one week after birth, showed the best performance, particularly in the XGB model. Its integration into treatment and management protocols can potentially reduce the incidence of permanent brain injury caused by IVH during the early stages of birth.

Abnormal amygdala volume moderates parenting and anxiety symptoms in children and adolescents with anxiety disorder

Anxiety disorders (AD) usually onset in childhood or adolescence and are related to brain development and early experiences during this period. As the hub of the fear circuit, the amygdala plays a crucial role in the development of emotional processing, and abnormalities in its structure and function are associated with anxiety disorders. We aim to uncover the amygdala volume's moderation between parenting and anxiety severity in children and adolescents with AD. 129 children and adolescents with anxiety and 135 age- and sex-matched Health controls (HC) using the publicly available Healthy Brain Network (HBN) dataset were included. Anxiety severity was measured using the Screen for Child Anxiety Related Disorders Self-report (SCARED-SR) and parenting was measured using the Alabama Parenting Questionnaire Self-Report (APQ-SR). We investigated age-related differences in amygdala volume in children and adolescents with anxiety disorders. Further, we examined the role of amygdala volume in moderating the association between parental involvement, particularly the maternal involvement, and anxiety symptoms in this population. We found larger bilateral amygdala in the AD group compared with the HC among the age range of 7–12. And increases in amygdala volume tended to negatively moderate the linear relationships between maternal involvement and anxiety symptoms in the AD group. These findings provide new evidence of abnormal brain alteration in children and adolescents with anxiety and may reflect proactive adaptations of adolescent brain development.

Reinforcement learning vs rule-based dynamic movement strategies in UAV assisted networks

Since resource allocation of cellular networks is not dynamic, some cells may experience unplanned high traffic demands due to unexpected events. Unmanned aerial vehicles (UAV) can be used to provide the additional bandwidth required for data offloading.Considering real-time and non-real-time traffic classes, our work is dedicated to optimize the placement of UAVs in cellular networks by two approaches. A first rule-based, low complexity method, that can be embedded in the UAV, while the other approach uses Reinforcement Learning (RL). It is based on Markov Decision Processes (MDP) for providing optimal results. The energy of the UAV battery and charging time constraints have been taken into account to cover a typical cellular environment consisting of many cells.We used an open dataset for the Milan cellular network provided by Telecom Italia to evaluate the performance of both proposed models. Considering this dataset, the MDP model outperforms the rule-based algorithm. Nevertheless, the rule-based one requires less processing complexity and can be used immediately without any prior data. This work makes a notable contribution to developing practical and optimal solutions for UAV deployment in modern cellular networks.

A typology of social network interactions in sub-Saharan Africa: Evidence from a rural population in Senegal

Social isolation/marginalization in sub-Saharan Africa is under-researched, despite increasing evidence of weakening traditional community-based social support. This paper aims to develop a typology of social networks capable of accounting for social marginalization in a rural community in Western Senegal and to describe the socio-demographic characteristics of network profiles. Building on prior qualitative work, we carry out a latent profile analysis using a unique and extensive social network data set, identifying four different network profiles: Locally integrated, Constrained relationships, Locally marginalized, and Local elites. This paper provides the first empirically supported classification of social integration and marginalization in social networks in rural sub-Saharan Africa. In doing so, it can serve as a reference for future research seeking to understand both the broader scope of social integration and marginalization and the consequences of differential access to social capital through social networks on access to health resources and well-being.