Computing最新文献

Heterogeneous graph representation learning has strong expressiveness when dealing with large-scale relational graph data, and its purpose is to effectively represent the semantic information and heterogeneous structure information of nodes in the graph. Current methods typically use shallow models to embed semantic information on low-order neighbor nodes in the graph, which prevents the complete retention of higher-order semantic feature information. To address this issue, this paper proposes a heterogeneous graph network for higher-order semantic enhancement called HHSE. Specifically, our model uses the identity mapping mechanism of residual attention at the node feature level to enhance the information representation of nodes in the hidden layer, and then utilizes two aggregation strategies to improve the retention of high-order semantic information. The semantic feature level aims to learn the semantic information of nodes in various meta path subgraphs. Extensive experiments on node classification and node clustering on three real-existing datasets show that the proposed approach makes practical improvements compared to the state-of-the-art methods. Besides, our method is applicable to large-scale heterogeneous graph representation learning.

Abstract

In this study, we propose a novel methodology called Particle Dynamics Method (PDM) for identifying and quantifying influential nodes in complex networks. Inspired by Newton’s three laws of motion and the universal gravitation law, PDM is based on a mathematical programming method that leverages node degrees and shortest path lengths. Unlike traditional centrality measures, PDM is easily adaptable to different network sizes and models, making it a versatile tool for network analysis. Our updated version of PDM also considers the direction of each force, resulting in more reliable results. To evaluate PDM’s performance, we tested it on a set of benchmark networks with distinct characteristics and models. Our results demonstrate that PDM outperforms other methodologies in the literature, as removing the identified influential nodes results in a significant decrease in network efficiency and robustness. The key feature of PDM is its flexibility in defining distance, which can be adapted to various network types. For instance, in a transportation network, distance can be defined by the flow between nodes, while in an academic publication system, the quartile of the journal could be used. Our research not only demonstrates the effectiveness of PDM but also highlights the influence of universities in the higher education and global university ranking networks, shedding light on the dynamics of these networks. Our interdisciplinary work has significant potential for collaborations between optimization, physics, and network science. This study opens up avenues for future research, including the extension of PDM to multilayer networks and the generalization of the metrics of monolayer networks for this purpose.

Abstract

To ensure the reliability and operational efficiency of the grid system, this paper proposes an unmanned aerial vehicle (UAV)-assisted Power Internet of Things (PIoT), which obtains real-time grid data through PIoT devices to support the management optimization of the grid system. Compared with traditional UAV-assisted communication networks, this paper enables data collection and energy transmission services for PIoT devices through UAVs. Firstly, the flight-hover-communication protocol is used. When the UAVs approach the target devices, they stop flying and remain hovering to provide services. The UAV selects full duplex mode in the hovering state, i.e., within the coverage area of the UAV, it can collect data from the target device while providing charging for other devices. Secondly, the UAVs can provide services to the required devices in sequence. Considering the priorities of the devices, both the data queue state and the energy pair state of network devices are considered comprehensively. Therefore, the optimization problem is constructed as a multi-objective optimization problem. First, the multi-objective optimization problem is transformed into a Markov process. Then, a multi-objective dynamic resource allocation algorithm based on reinforcement learning is proposed for solving the multi-objective optimization problem. The simulation results show that the proposed resource allocation scheme can effectively achieve a reasonable allocation of UAV resources, joint multi-objective optimization, and improved system performance.

Wireless Sensor Networks (WSNs) often operate in hostile environments and are subject to frequent failures. Failure of multiple sensor nodes causes the network to split into disjoint segments, which leads to network partitioning. Federating these disjoint segments is necessary to prevent detrimental effects on WSN applications. This paper investigates a recovery strategy using mobile relay nodes (MD-carrier) for restoring network connectivity. The proposed MD-carrier Tour Planning (MDTP) approach restores network connectivity of partitioned WSNs with reduced tour length and latency. For this reason, failure nodes are identified, and disjoint segments are formed with the k-means algorithm. Then, the Analytic Hierarchy Process (AHP) and the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) are used for the election of an AGgregator Node (AGN) for each segment. Furthermore, an algorithm for identifying sojourn locations is proposed, which coordinates the maximum number of AGNs. Choosing the sojourn locations is a challenging task in WSN since the incorrect selection of the sojourn locations would degrade its data collection process. This paper uses the nature-inspired meta-heuristic Donkey And Smuggler Optimization (DASO) algorithm to compute the optimal touring path. MDTP reduces tour length and latency by an average of 30.28% & 24.56% compared to existing approaches.

This paper presents an approach based on deep learning for accurate Electrocardiogram signal classification. The electrocardiogram is a significant signal in the realm of medical affairs, which gives vital information about the cardiovascular status of patients to heart specialists. Manually meticulous analysis of signals needs high and specific skills, and it is a time-consuming job too. The existence of noise, the inflexibility of signals, and the irregularity of heartbeats keep heart specialists in trouble. Cardiovascular diseases (CVDs) are the most important factor of fatality globally, which annually caused the deaths of 17.9 million people. Totally 31% of all death in the world are related to CVDs, which the age of 1/3 of patients that died because of CVDs is below 70 Because of the high percentage of mortality in cardiovascular patients, accurate diagnosis of this disease is an important matter. We present an approach to the analysis of electrocardiogram signals based on the convolutional neural network, discrete wavelet transformation with db2 mother wavelet, and synthetic minority over-sampling technique (SMOTE) on the MIT-BIH dataset according to the association for the advancement of medical instrumentation (AAMI) standards to increase the accuracy in electrocardiogram signal classifications. The evaluation results show this approach with 50 epoch training that the time of each epoch is 39 s, achieved 99.71% accuracy for category F, 98.69% accuracy for category N, 99.45% accuracy for category S, 99.33% accuracy for category V and 99.82% accuracy for category Q. It is worth mentioning that it can potentially be used as a clinical auxiliary diagnostic tool. The source code is available at https://gitlab.com/arminshoughi/ecg-classification-cnn.

Some computations are based on structures of random partition. They take an n-size problem as input, then break this problem into sub-problems of randomized size, execute calculations on each sub-problems and combine results from these calculations at last. We propose a combinatorial method for analyzing such computations and prove that the averaged time complexity is in terms of Stirling numbers of the second kind. The result shows that the average time complexity is decreased about one order of magnitude compared to that of the original solution. We also show two application cases where random partition structures are applied to improve performance.

The topological relationship of spatial data is essential to GIS data processing and spatial analysis such as in analysis of pipe explosion in gas pipeline network. The existing browser-side JavaScript topology check library is inefficient and even crashes when checking the pipe network topology relationships for large amounts of data. In this paper, we present a topology checking and optimization method for pipeline vector data in browser-side using quadtree. Firstly, an algorithm mechanism that conforms to GIS data is designed based on JavaScript shared memory mechanism, topological check algorithm characteristics, and spatial data high-precision characteristics. Then using a fast rejection experiment and straddle test to realize the browser-side topology checking algorithm, through tolerance setting, improve the inspection efficiency and accuracy, which solves the problem that Turf and Jsts libraries cannot set tolerance. Based on the concept of quadtree spatial index, an optimization method of browser-side quadtree topology checking algorithm(BQTCA) is proposed. Without setting tolerance, the topology check of 114 point data and 1881 line data takes 487 milliseconds, and the efficiency of BQTCA is about 12 times and 39 times higher than that of the well-known public libraries Turf and Jsts, respectively. When the data volume increases to 912 point data and 15048 line data, BQTCA takes 6970 ms, which is about 65 times and 190 times more efficient than Turf and Jsts, respectively. The larger the data volume is, the more pronounced the efficiency improvement of BQTCA. Even when the data volume is so large that Turf and Jsts can- not calculate even crash, BQTCA can still complete the checking calculation. Through experiments, BQTCA can significantly improve the efficiency of browser-side vector pipeline topology relationship inspection under a large amount of data, and meet the commercial application requirements.

At present, large-scale sky surveys have obtained a large volume of stellar spectra. An efficient classification algorithm is of great importance to the practice of astronomical research. In this paper, we propose a novel parallel optimization algorithm based on a lock-free and shared-memory environment to solve the model for astronomical spectra class. Firstly, the SMOTE-TOMEK and RobustScaler are introduced to use for class balancing and data normalization. Secondly, 1-Dimensional Convolutional Neural Networks (1-D CNN) with L2-norm loss function is utilized as a classifier. Finally, LFA-SGD, LFA-Adagrad, LFA-RMSprop and LFA-Adam algorithms are proposed and applied to the classifier solution. The Lock-Free and shared-memory parallel Asynchronous environment (LFA) relies on GPU multiprocessing, allowing the algorithm to fully utilize the multi-core resources of the computer. Due to its sparsity, the convergence speed is significantly faster. The experimental results show that LFA-SGD algorithm and its variants achieved state-of-the-art accuracy and efficiency for astronomical spectra class.