Current Graph Research Articles

Early Forecast of Traffic Accident Impact Based on a Single-Snapshot Observation (Student Abstract)

Predicting and quantifying the impact of traffic accidents is necessary and critical to Intelligent Transport Systems (ITS). As a state-of-the-art technique in graph learning, current graph neural networks heavily rely on graph Fourier transform, assuming homophily among the neighborhood. However, the homophily assumption makes it challenging to characterize abrupt signals such as traffic accidents. Our paper proposes an abrupt graph wavelet network (AGWN) to model traffic accidents and predict their time durations using only one single snapshot.

Personalized query techniques in graphs: A survey

Graph is a famous data structure that has prevalent applications in the real world, including social networks, biological networks, and computer networks. In these applications, graph management operators are powerful tools for mining important information hidden in large-scale graphs. As important graph data management operators, personalized graph queries are playing an increasingly significant role in providing users with effective decision support. In particular, the purpose of personalized graph queries is to compute personalized results which can meet the preferences of different users from the three aspects of specified query vertices, structures, and attributes. In this paper, we conduct a survey to offer a comprehensive view of the current personalized graph queries which need users to specify query vertices over simple and attributed graphs, respectively. These queries fall into three categories, including point-related, path-related, and subgraph-related graph queries, whose query results have distinct structures. We analyze existing approaches to personalized graph queries and highlight current challenges. In addition, we also offer guidelines for future graph queries.

News topic detection based on capsule semantic graph

Most news topic detection methods use word-based methods, which easily ignore the relationship among words and have semantic sparsity, resulting in low topic detection accuracy. In addition, the current mainstream probability methods and graph analysis methods for topic detection have high time complexity. For these reasons, we present a news topic detection model on the basis of capsule semantic graph (CSG). The keywords that appear in each text at the same time are modeled as a keyword graph, which is divided into multiple subgraphs through community detection. Each subgraph contains a group of closely related keywords. The graph is used as the vertex of CSG. The semantic relationship among the vertices is obtained by calculating the similarity of the average word vector of each vertex. At the same time, the news text is clustered using the incremental clustering method, where each text uses CSG; that is, the similarity among texts is calculated by the graph kernel. The relationship between vertices and edges is also considered when calculating the similarity. Experimental results on three standard datasets show that CSG can obtain higher precision, recall, and F1 values than several latest methods. Experimental results on large-scale news datasets reveal that the time complexity of CSG is lower than that of probabilistic methods and other graph analysis methods.

Development of mobile application for wireless sensor networks for efficient irrigation water management

The evolution of smart phones has necessitated the development of mobile applications designed to perform a wide variety of functions. In the field of agriculture, mobile applications are currently used to monitor environmental parameters such as ambient temperature, humidity, soil moisture, water level, among others. A mobile application intended to monitor irrigation-related parameters and to control solenoid valves for irrigation automation was developed in this study. The mobile application was written using Flutter software development kit, and the Dart programming language. The mobile application communicates with the cloud server using a REST API written in JavaScript. The data acquired from the cloud server are presented as the current sensor reading and graphs. On the other hand, the mobile application controls the solenoid valves by sending designated bytes of data to the cloud server. The mobile application developed in this study was designed to be integrated with both low-cost sensors and the Smartmesh IP sensors to enable real-time monitoring and data visualization, and facilitate irrigation scheduling and manual irrigation control. The mobile application developed in this study may be used for efficient irrigation water management of upland crop production systems and for agricultural modernization in the Philippines and other developing countries.

Time Series Prediction of Sea Surface Temperature Based on an Adaptive Graph Learning Neural Model

The accurate prediction of sea surface temperature (SST) is the basis for our understanding of local and global climate characteristics. At present, the existing sea temperature prediction methods fail to take full advantage of the potential spatial dependence between variables. Among them, graph neural networks (GNNs) modeled on the relationships between variables can better deal with space–time dependency issues. However, most of the current graph neural networks are applied to data that already have a good graph structure, while in SST data, the dependency relationship between spatial points needs to be excavated rather than existing as prior knowledge. In order to predict SST more accurately and break through the bottleneck of existing SST prediction methods, we urgently need to develop an adaptive SST prediction method that is independent of predefined graph structures and can take full advantage of the real temporal and spatial correlations hidden indata sets. Therefore, this paper presents a graph neural network model designed specifically for space–time sequence prediction that can automatically learn the relationships between variables and model them. The model automatically extracts the dependencies between sea temperature multi-variates by embedding the nodes of the adaptive graph learning module, so that the fine-grained spatial correlations hidden in the sequence data can be accurately captured. Figure learning modules, graph convolution modules, and time convolution modules are integrated into a unified end-to-end framework for learning. Experiments were carried out on the Bohai Sea surface temperature data set and the South China Sea surface temperature data set, and the results show that the model presented in this paper is significantly better than other sea temperature model predictions in two remote-sensing sea temperature data sets and the surface temperature of the South China Sea is easier to predict than the surface temperature of the Bohai Sea.

Federated Dynamic Graph Neural Networks with Secure Aggregation for Video-based Distributed Surveillance

Distributed surveillance systems have the ability to detect, track, and snapshot objects moving around in a certain space. The systems generate video data from multiple personal devices or street cameras. Intelligent video-analysis models are needed to learn dynamic representation of the objects for detection and tracking. Can we exploit the structural and dynamic information without storing the spatiotemporal video data at a central server that leads to a violation of user privacy? In this work, we introduce Federated Dynamic Graph Neural Network (Feddy), a distributed and secured framework to learn the object representations from graph sequences: (1) It aggregates structural information from nearby objects in the current graph as well as dynamic information from those in the previous graph. It uses a self-supervised loss of predicting the trajectories of objects. (2) It is trained in a federated learning manner. The centrally located server sends the model to user devices. Local models on the respective user devices learn and periodically send their learning to the central server without ever exposing the user’s data to server. (3) Studies showed that the aggregated parameters could be inspected though decrypted when broadcast to clients for model synchronizing, after the server performed a weighted average. We design an appropriate aggregation mechanism of secure aggregation primitives that can protect the security and privacy in federated learning with scalability. Experiments on four video camera datasets as well as simulation demonstrate that Feddy achieves great effectiveness and security.

Brain Functional Connectivity Analysis via Graphical Deep Learning.

Graphical deep learning models provide a desirable way for brain functional connectivity analysis. However, the application of current graph deep learning models to brain network analysis is challenging due to the limited sample size and complex relationships between different brain regions. In this work, a graph convolutional network (GCN) based framework is proposed by exploiting the information from both region-to-region connectivities of the brain and subject-subject relationships. We first construct an affinity subject-subject graph followed by GCN analysis. A Laplacian regularization term is introduced in our model to tackle the overfitting problem. We apply and validate the proposed model to the Philadelphia Neurodevelopmental Cohort for the brain cognition study. Experimental analysis shows that our proposed framework outperforms other competing models in classifying groups with low and high Wide Range Achievement Test (WRAT) scores. Moreover, to examine each brain region's contribution to cognitive function, we use the occlusion sensitivity analysis method to identify cognition-related brain functional networks. The results are consistent with previous research yet yield new findings. Our study demonstrates that GCN incorporating prior knowledge about brain networks offers a powerful way to detect important brain networks and regions associated with cognitive functions.

LighterKGCN: A Recommender System Model based on Bi-layer Graph Convolutional Networks

<p>Recommender systems have been extensively utilized to meet users’ personalized needs. Collaborative filtering is one of the most classic algorithms in the recommendation field. However, it has problems such as cold start and data sparsity. In that case, knowledge graphs and graph convolutional networks have been introduced by scholars into recommender systems to solve the above problems. However, the current graph convolutional networks fail to give full play to the advantages of graph convolution since they are employed either in the embedding representations of users and commodity entities, or in the embedding representations between entities of the knowledge graphs. Therefore, LighterKGCN, a recommender system model based on bi-layer graph convolutional networks was proposed in accordance with the KGCN model and the LightGCN model. In the first layer of GCN, the model first learned the embedding representations of users and commodity entities on the user-commodity entity interaction graph. Then, the attained user embedding and commodity embedding were used as the data source for the second layer of GCN. In the second layer, the entity v and its neighborhoods were calculated using the hybrid aggregation function proposed in this paper. The result was taken as the new entity v. According to tests on three public datasets and comparison results with the KGCN, LighterKGCN improved by 0.52% and 51.16% in terms of AUC and F1 performances, respectively on the dataset of MovieLens-20M; LighterKGCN improved by 0.67% and 45.0% in terms of AUC and F1 performances, respectively on the dataset of Yelp2018; and the number was 0.67% and 36.35% in AUC and F1 performances, respectively on the dataset of Last.FM.</p> <p> </p>

A simple proof of the Map Color Theorem for nonorientable surfaces

We give a more simple proof of the Map Color Theorem for nonorientable surfaces that uses only four constructions of current graphs instead of 12 constructions used in the previous proof. For every i=0,1,2,3, using an index one current graph with cyclic current group, we construct a nonorientable triangular embedding of K12s+3i+1 that can be easily modified into a nonorientable triangular embedding of K12s+3i+3 and a minimal nonorientable embedding of K12s+3i+5. As a result, for every n≥10, n∉{11,14,20}, we construct a minimal nonorientable embedding of Kn. During the modifications, all but 2(4s+i) faces of the nonorientable triangular embedding of K12s+3i+1 become faces of the nonorientable triangular embedding of K12s+3i+3, and all but 2(4s+i)+1 faces of the nonorientable triangular embedding of K12s+3i+3 become faces of the minimal nonorientable embedding of K12s+3i+5.

A Novel Double-layer Low-profile Multiband Frequency Selective Surface for 4G Mobile Communication System

A novel double-layer multiband, low-profile frequency selective surface (FSS) for IMT-Advanced (4G) mobile communication system is presented in this article. On aspired to a minimum transmission coefficient of −10 dB for surface materials when the frequency bands targeted for blocking are stopped. For this project, we chose the dielectric substrate FR4 (loss-tangent = 0.02; dielectric constant = 4.54) and a thickness of 1 mm. Dodecagonal rings, upright bars, and square frame make up the FSS unit cell. The desired frequency responses of the FSS were intended to avoid being changed according to the angle of incidence of the electromagnetic waves. The FSS design is proposed as a symmetrical structure to make it polarization-independent and is aimed to stop 800, 900, 1800, 2100, and 2600 MHz frequencies to prevent harmful effects to human health and interference effects at these frequencies. With a cell size of 0.17λ, the planned FSS is quite small and, thus, has a low sensitivity at the angle of the incident wave. In addition, FSS geometry was manufactured by a printed circuit board (PCB) and measured in a non-reflective environment after being studied in Ansys high-frequency structure simulator (HFSS) software. By comparing the analysis and measurement results of the design, the success of the FSS to the frequencies to be stopped has been verified. The effect of each patch on different frequencies has been examined by drawing the surface current density graphs of the design.

Graph regression for pressure peak prediction in fracturing processes

In the oil and gas industrial field, fracturing construction technology is commonly used for increasing oil–gas production. A major concern in fracturing construction is whether the pressure at the wellhead will exceed safety threshold, when a large amount of sand-containing fluid is periodically injected into a well. To avoid accidents like blowout caused by extreme wellhead pressure, it is critically important to monitor pressure change in real time. In this work we propose to predict the pressure peaks during each fracturing period, for better designing future fracturing strategy. Towards this end, we present a novel non-parametric graph regression method, which is able to model the correlation of historical pressure peaks and learn the features of fracturing cycles via Laplacian-smoothness based graph learning. whereby the periodic fracturing signals, namely, fracking fluid concentration sequence, together with peaks, are able to be encoded in a latent Euclidean space by Laplacian-smoothness inspired graph learning. Then we introduce non-parametric linear regression for peak prediction based on the most similar nodes (e.g. peaks) w.r.t the node in the predicting period. Meanwhile, the current graph is updated using the predicted peak value. In the graph update, we particularly introduce a node forgetting mechanism to control the graph scale and to reduce the computational complexity, so as to achieve rapid prediction of construction operations. We conduct extensive experiments on real-world datasets from a well. The experimental results demonstrate that the proposed method can effectively predict the oil pressure peak and significantly outperform the state-of-the-art models.

Graft: A graph based time series data mining framework

Rapid technology integration causes a high dimensional time series data accumulation in multiple domains and applying the classical data mining tools and techniques becomes a challenging task. Hence, the time series data representation have gained popularity over the years, which ease the task of mining, analysis and visualization. Graph based representation is one such emerging tool in which the time series data is represented as nodes and edges of graph. The current graph based representation is designed either to mine the motif or discords from a single time series or cluster the time series where each node represents a time series sample. Such representation technique causes information loss and also no further analysis could be performed other than clustering. To address these challenges, we propose a unique graph representation for time series dataset that works on multiple domains. Novelty of the graph representation is that it is unique for multiple time series and it acts as a framework for whole time series clustering, temporal pattern extraction from each cluster and temporally dependent rare event discovery. A new research direction for the proposed graph based framework is shown. Comparative analysis reveal the superiority of the proposed framework particularly as a clustering technique. The key contributions of the paper are: (i) transformation strategy of time series database from time domain to graph structure in topological domain (ii) time series clustering using path level analysis (iii) identification of temporally dependent co-occurring patterns (iv) rare event detection using component level analysis

Connectogram – A graph-based time dependent representation for sounds

The proposed method contributes the time-series classification literature with a novel time-convexity based representation, which extends the current graph conversion approaches by introducing the time dimension, also introducing a colorful graph-generator approach. The representation capability of connectograms is tested in comparison with mel-spectrograms (mels) and MFCCs for an environmental sound classification task, as input to state-of-art transfer learning models. Results indicate that connectograms cannot compete with the best-performer mel-spectrogram representations in standalone format, however they significantly improve their classification performance in case they are combined as single layers of hybrid RGB representations. A combination of [mels + mels + connectogram] outperforms either sole representations or their combinations by 2–3%, with 96.46% classification accuracy for ResNet50 classifier model.

Towards PageRank Update in a Streaming Graph by Incremental Random Walk

As the internet and the Internet of Things (IoT) have been widely applied in many application fields, a large number of graphs are continuously produced and change over time, which leads to difficulties in graph analysis and utilization. This paper studies a PageRank update algorithm for a streaming graph using incremental random walk. We focus on the information about the local changes of nodes and edges in the current graph, analyze the impact of such local changes on this current graph, and then use the idea behind wave propagation theory to seek and determine all affected nodes that need to be updated their PageRank in the new graph. For new nodes, the existing nodes in the current graph that are connected with these new nodes are aggregated into a supernode, and the PageRank of the new nodes is solved in the new graph with the supernode. Finally, we conduct a series of experiments on real-world and synthetic graph datasets. Compared with the state-of-the-art incremental computing algorithm, our algorithm not only ensures the accuracy of calculating the PageRank in a large streaming graph, but also speeds up the computational process by avoiding many redundant computations.

Boosting Climate Analysis With Semantically Uplifted Knowledge Graphs

Nowadays, the fast expansion of heterogeneous climate data resources accessible on the Internet has led to substantial data fragmentation on the web. For example, station-based sensor data from different sources are likely to be interrelated but may be stored in disparate formats, such as <monospace>CSV</monospace>, <monospace>JSON</monospace>, and <monospace>XML</monospace>. To address the data isolation problem, several semantically uplifted knowledge graphs are proposed for climate data exchange. While these knowledge graphs improve data interoperability, the advancement in multisource data interchange is limited to data included inside knowledge graphs. As a result, the exclusive interoperability of current climatic knowledge graphs hampers the flow of data into typical climate analysis workflows in contexts, where analytical models often need data in nonknowledge graph formats. This article addresses this issue by focusing on enhancing climate analysis workflows within the context of the Python machine learning environment, with a preference for tabular data. We propose an analysis workflow able to automatically integrate remote climate knowledge graph data with local tabular data so as to enhance the data usability with respect to certain climate analysis tasks. To underscore the importance of our study, we illustrate how the workflow streamlines the access to multisource climatic variables in the Python environment from a semantic perspective. The additional knowledge graph data have the potential to augment local datasets in the climate domain, as evidenced by an improvement in accuracy of up to 10&#x0025; for machine learning geared on rainfall detection.

Knowledge graph embedding based on semantic hierarchy

In view of the current knowledge graph embedding, it mainly focuses on symmetry/opposition, inversion and combination of relationship patterns, and does not fully consider the structure of the knowledge graph. We propose a Knowledge Graph Embedding Based on Semantic Hierarchy (SHKE), which fully considers the information of knowledge graph by fusing the semantic information of the knowledge graph and the hierarchical information. The knowledge graph is mapped to a polar coordinate system, where concentric circles naturally reflect the hierarchy, and entities can be divided into modulus parts and phase parts, and then the modulus part of the polar coordinate system is mapped to the relational vector space through the relational vector, thus the modulus part takes into account the semantic information of the knowledge graph, and the phase part takes into account the hierarchical information. Experiments show that compared with other models, the proposed model improves the knowledge graph link prediction index Hits@10% by about 10% and the accuracy of the triple group classification experiment by about 10%.

Attention Regularized Laplace Graph for Domain Adaptation.

In leveraging manifold learning in domain adaptation (DA), graph embedding-based DA methods have shown their effectiveness in preserving data manifold through the Laplace graph. However, current graph embedding DA methods suffer from two issues: 1). they are only concerned with preservation of the underlying data structures in the embedding and ignore sub-domain adaptation, which requires taking into account intra-class similarity and inter-class dissimilarity, thereby leading to negative transfer; 2). manifold learning is proposed across different feature/label spaces separately, thereby hindering unified comprehensive manifold learning. In this paper, starting from our previous DGA-DA, we propose a novel DA method, namely A ttention R egularized Laplace G raph-based D omain A daptation (ARG-DA), to remedy the aforementioned issues. Specifically, by weighting the importance across different sub-domain adaptation tasks, we propose the A ttention R egularized Laplace Graph for class aware DA, thereby generating the attention regularized DA. Furthermore, using a specifically designed FEEL strategy, our approach dynamically unifies alignment of the manifold structures across different feature/label spaces, thus leading to comprehensive manifold learning. Comprehensive experiments are carried out to verify the effectiveness of the proposed DA method, which consistently outperforms the state of the art DA methods on 7 standard DA benchmarks, i.e., 37 cross-domain image classification tasks including object, face, and digit images. An in-depth analysis of the proposed DA method is also discussed, including sensitivity, convergence, and robustness.

Задача локальної оптимізації графів та її застосування

Introduction. We study the problem of constructing a locally optimal directed graph when each edge is assigned a fixed numerical weight. Due to the broad interpretation of the numerical weights of edges, the problem of finding the least-cost paths has many applications and is one of the most common and studied problems in the applied theory of algorithms. In various modifications, it can be found in algorithms for constructing optimal routes for flying and other moving objects, in pattern recognition, optimization of communication networks and integrated circuits. The property of local optimality is defined through a dynamic process that at every round assigns to each vertex the minimal value depending on its current value and sums of values stored in neighborhood vertices with weights of incoming edges. The graph is locally optimal when any vertex cannot allow further minimizations. Purpose. The purpose of the article is to investigate the problem of constructing a locally-optimal graph, establish properties of such graphs, and derive new applications. Methods. We consider the problem of finding the shortest paths in a broader context as a general computational problem of local graph optimization. This model of computations on a graph is studied. Each local vertex, regardless of the other vertices, performs the same iterative procedure of minimizing some non-negative numerical value assigned to that vertex. The operation of such minimization depends on the current value obtained in this vertex when applying the previous minimization and sums of the current numerical data located in the vertices from which the input edges lead to the selected local vertex and weights of the seedges. After several stages of such an operation, the graph always comes into a stable state: further minimization of vertices does not change the values assigned to the vertices. In this state, we call the graph locally optimal. Results. The conditions under which the graph reaches the state of local optimization are investigated. With different choices of interpretation of the function that determines the influence of neighboring vertices on the vertex minimization value, we obtain a variety of applications. For example, the gradient descent method from the local optimization state makes it easy to obtain the minimum weight (cost) paths from the given starting vertex to all vertices. If the graph is interpreted as a semantic network in the space of natural language words, it is shown how the problem of finding the minimum path among words or phrases is interpreted as solving a known problem in computational linguistics: find the word most associated with a given input set words or sentences (images). Finally, by calculating the weights of the paths in randomized (uniformly distributed) graphs, a fast-authentication protocol is proposed for the two coalition entities acting in a malicious environment. Conclusions. The local optimality is a stable constant state of the current graph marking reachable by using the dynamic minimization process. Local optimality provides a common method for solving many problems related to search engine optimization. It is shown how from solving the general problem of constructing a locally-optimal graph it follows the problem of finding the paths of least weights (cost). New applications include solving the problem of associative search in computational linguistics (creative thinking) and fast mutual authentication in coalition groups.

COMPARATIVE ANALYSIS OF SIX PHASE INDUCTION MOTOR

In recent years, AC induction motors have replaced DC motor. The most important reason for this is that DC motors have a mechanical brush and commutator system that is subject to wear and tear over time. While DC motor drivers need a lot of maintenance due to this system, induction motors do not require almost any maintenance since this system is not available. In recent years, technological advances in the field of multiphase induction motors have led to a significant evolution. A lot of research has been done around the world and a lot of new and interesting developments have been reported to the literature. This paper provides information about six-phase induction. In addition, what kind of advantages the concept of multiphase provides and how this structure is formed contains information about them. The six-phase induction motor will be simulated through the MATLAB/SIMULINK package program and compare with three phase induction motor. Both motors were run in simulation using the same parameters. At the end of the simulation, current, moment and speed graphs of the six-phase induction motor and three-phase induction motor will be drawn according to time. The study also included equations used during modeling and equivalent circuits belonging to a six-phase induction motor. It has been graphically shown that a six-phase induction motor is more advantageous than a three-phase motor. For example, in six-phase induction motors, the transition time to a continuous state is shorter. In addition, while the frequency of torque vibration increases, its amplitude decreases, the current value of each phase decreases, the stator copper losses are less than the three-phase induction motor. In motors, usually the power needed is divided by the number of phases. So the more phases there are, the greater the power carrying capacity. Today, multi-phase induction motor drives are used in special applications such as electric vehicles, space technology, aerospace applications, as moving power on ships, electric trains, which require high reliability and power.

Semi-supervised node classification via adaptive graph smoothing networks

Inspections on current graph neural networks suggest us to reconsider the computational aspect of the final aggregation. We consider that such aggregations perform a prediction smoothing and impute their potential drawbacks to be the inter-class interference implied by the underlying graphs. We aim at weakening the inter-class connections so that aggregations focus more on intra-class relations and producing smooth predictions according to weakening results. We apply a metric learning module to learn new edge weights and combine entropy losses to ensure the correspondence between the predictions and the learnt distances so that the weights of inter-class edges are reduced and predictions are smoothed according to the modified graph. Experiments on four citation networks and a Wiki network show that in comparison with other state-of-the-art graph neural networks, the proposed algorithm can improve the classification accuracy.