Single Graph Research Articles

Situational Graphs for Robot Navigation in Structured Indoor Environments

Mobile robots should be aware of their <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">situation</i> , comprising the deep understanding of their surrounding environment along with the estimation of its own state, to successfully make intelligent decisions and execute tasks autonomously in real environments. 3D scene graphs are an emerging field of research that propose to represent the environment in a joint model comprising geometric, semantic and relational/topological dimensions. Although 3D scene graphs have already been combined with SLAM techniques to provide robots with situational understanding, further research is still required to effectively deploy them on-board mobile robots. To this end, we present in this paper a novel, real-time, online built Situational Graph ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">S-Graph</i> ), which combines in a single optimizable graph, the representation of the environment with the aforementioned three dimensions, together with the robot pose. Our method utilizes odometry readings and planar surfaces extracted from 3D LiDAR scans, to construct and optimize in real-time a three layered <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">S-Graph</i> that includes (1) a robot tracking layer where the robot poses are registered, (2) a metric-semantic layer with features such as planar walls and (3) our novel topological layer constraining the planar walls using higher-level features such as corridors and rooms. Our proposal does not only demonstrate state-of-the-art results for pose estimation of the robot, but also contributes with a metric-semantic-topological model of the environment.

Attraction and Repulsion: Unsupervised Domain Adaptive Graph Contrastive Learning Network

Graph convolutional networks (GCNs) are important techniques for analytics tasks related to graph data. To date, most GCNs are designed for a single graph domain. They are incapable of transferring knowledge from/to different domains (graphs), due to the limitation in graph representation learning and domain adaptation across graph domains. This paper proposes a novel Graph Contrastive Learning Network (GCLN) for unsupervised domain adaptive graph learning. The key innovation is to enforce attraction and repulsion forces within each single graph domain, and across two graph domains. Within each graph, an attraction force encourages local patch node features to be similar to global representation of the entire graph, whereas a repulsion force will repel node features so they can separate network from its permutations ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e.</i> domain-specific graph contrastive learning). Across two graph domains, an attraction force encourages node features from two domains to be largely consistent, whereas a repulsion force ensures features are discriminative to differentiate graph domains ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e.</i> cross-domain graph contrastive learning). The within- and cross-domain graph contrastive learning is carried out by optimizing an objective function, which combines source classifier and target classifier loss, domain-specific contrastive loss, and cross-domain contrastive loss. As a result, feature learning from graphs is facilitated using knowledge transferred between graphs. Experiments on real-world datasets demonstrate that GCLN outperforms state-of-the-art graph neural network algorithms.

The druggable genome: Twenty years later.

The concept of the druggable genome has been with us for 20 years. During this time, researchers have developed several methods and resources to help assess a target’s druggability. In parallel, evidence for target-disease associations has been collated at scale by Open Targets. More recently, the Protein Data Bank in Europe (PDBe) have built a knowledge base matching per-residue annotations with available protein structure. While each resource is useful in isolation, we believe there is enormous potential in bringing all relevant data into a single knowledge graph, from gene-level to protein residue. Automation is vital for the processing and assessment of all available structures. We have developed scalable, automated workflows that provide hotspot-based druggability assessments for all available structures across large numbers of targets. Ultimately, we will run our method at a proteome scale, an ambition made more realistic by the arrival of AlphaFold 2. Bringing together annotations from the residue up to the gene level and building connections within the graph to represent pathways or protein-protein interactions will create complexity that mirrors the biological systems they represent. Such complexity is difficult for the human mind to utilise effectively, particularly at scale. We believe that graph-based AI methods will be able to expertly navigate such a knowledge graph, selecting the targets of the future.

Cohen-Macaulay Growing Graphs

We introduce a new family of simple graphs, so called, growing graphs. We investigate ways to modify a given simple graph G combinatorially to obtain a growing graph. One may obtain infinitely many growing graphs from a single simple graph. We show that a growing graph obtained from any given simple graph is Cohen–Macaulay and every Cohen–Macaulay chordal graph is a growing graph. We also prove that under certain conditions, a graph is growing if and only if its clique complex is grafted and give several equivalent conditions in this case. Our work is inspired by and generalizes a result of Villarreal on the use of whiskers and the work of Faridi on grafting of simplicial complexes.

Using the Information Harm Triangle to Identify Risk-Informed Cybersecurity Strategies for Instrumentation and Control Systems

The Information Harm Triangle (IHT) is a novel approach that aims to adapt intuitive engineering concepts to simplify defense in depth for instrumentation and control (I&C) systems at nuclear power plants. This approach combines digital harm, real-world harm, and unsafe control actions (UCAs) into a single graph named “Information Harm Triangle.” The IHT is based on the postulation that the consequences of cyberattacks targeting I&C systems can be expressed in terms of two orthogonal components: a component representing the magnitude of data harm (DH) (i.e., digital information harm) and a component representing physical information harm (PIH) (i.e., real-world harm, e.g., an inadvertent plant trip). The magnitude of the severity of the physical consequence is the aspect of risk that is of concern. The sum of these two components represents the total information harm. The IHT intuitively informs risk-informed cybersecurity strategies that employ independent measures that either act to prevent, reduce, or mitigate DH or PIH. Another aspect of the IHT is that the DH can result in cyber-initiated UCAs that result in severe physical consequences. The orthogonality of DH and PIH provides insights into designing effective defense in depth. The IHT can also represent cyberattacks that have the potential to impede, evade, or compromise countermeasures from taking appropriate action to reduce, stop, or mitigate the harm caused by such UCAs. Cyber-initiated UCAs transform DH to PIH.

Highly scalable maximum likelihood and conjugate Bayesian inference for ERGMs on graph sets with equivalent vertices.

The exponential family random graph modeling (ERGM) framework provides a highly flexible approach for the statistical analysis of networks (i.e., graphs). As ERGMs with dyadic dependence involve normalizing factors that are extremely costly to compute, practical strategies for ERGMs inference generally employ a variety of approximations or other workarounds. Markov Chain Monte Carlo maximum likelihood (MCMC MLE) provides a powerful tool to approximate the maximum likelihood estimator (MLE) of ERGM parameters, and is generally feasible for typical models on single networks with as many as a few thousand nodes. MCMC-based algorithms for Bayesian analysis are more expensive, and high-quality answers are challenging to obtain on large graphs. For both strategies, extension to the pooled case—in which we observe multiple networks from a common generative process—adds further computational cost, with both time and memory scaling linearly in the number of graphs. This becomes prohibitive for large networks, or cases in which large numbers of graph observations are available. Here, we exploit some basic properties of the discrete exponential families to develop an approach for ERGM inference in the pooled case that (where applicable) allows an arbitrarily large number of graph observations to be fit at no additional computational cost beyond preprocessing the data itself. Moreover, a variant of our approach can also be used to perform Bayesian inference under conjugate priors, again with no additional computational cost in the estimation phase. The latter can be employed either for single graph observations, or for observations from graph sets. As we show, the conjugate prior is easily specified, and is well-suited to applications such as regularization. Simulation studies show that the pooled method leads to estimates with good frequentist properties, and posterior estimates under the conjugate prior are well-behaved. We demonstrate the usefulness of our approach with applications to pooled analysis of brain functional connectivity networks and to replicated x-ray crystal structures of hen egg-white lysozyme.

Bot-MGAT: A Transfer Learning Model Based on a Multi-View Graph Attention Network to Detect Social Bots

Twitter, as a popular social network, has been targeted by different bot attacks. Detecting social bots is a challenging task, due to their evolving capacity to avoid detection. Extensive research efforts have proposed different techniques and approaches to solving this problem. Due to the scarcity of recently updated labeled data, the performance of detection systems degrades when exposed to a new dataset. Therefore, semi-supervised learning (SSL) techniques can improve performance, using both labeled and unlabeled examples. In this paper, we propose a framework based on the multi-view graph attention mechanism using a transfer learning (TL) approach, to predict social bots. We called the framework ‘Bot-MGAT’, which stands for bot multi-view graph attention network. The framework used both labeled and unlabeled data. We used profile features to reduce the overheads of the feature engineering. We executed our experiments on a recent benchmark dataset that included representative samples of social bots with graph structural information and profile features only. We applied cross-validation to avoid uncertainty in the model’s performance. Bot-MGAT was evaluated using graph SSL techniques: single graph attention networks (GAT), graph convolutional networks (GCN), and relational graph convolutional networks (RGCN). We compared Bot-MGAT to related work in the field of bot detection. The results of Bot-MGAT with TL outperformed, with an accuracy score of 97.8%, an F1 score of 0.9842, and an MCC score of 0.9481.

Properties of graphs specified by a regular language

Traditionally, graph algorithms get a single graph as input, and then they should decide if this graph satisfies a certain property varPhi . What happens if this question is modified in a way that we get a possibly infinite family of graphs as an input, and the question is if there is a graph satisfying varPhi in the family? We approach this question by using formal languages for specifying families of graphs, in particular by regular sets of words. We show that certain graph properties can be decided by studying the syntactic monoid of the specification language L if a certain torsion condition is satisfied. This condition holds trivially if L is regular. More specifically, we use a natural binary encoding of finite graphs over a binary alphabet varSigma , and we define a regular set mathbb {G}subseteq varSigma ^* such that every nonempty word win mathbb {G} defines a finite and nonempty graph. Also, graph properties can then be syntactically defined as languages over varSigma . Then, we ask whether the automaton mathcal {A} specifies some graph satisfying a certain property varPhi . Our structural results show that we can answer this question for all “typical” graph properties. In order to show our results, we split L into a finite union of subsets and every subset of this union defines in a natural way a single finite graph F where some edges and vertices are marked. The marked graph in turn defines an infinite graph F^infty and therefore the family of finite subgraphs of F^infty where F appears as an induced subgraph. This yields a geometric description of all graphs specified by L based on splitting L into finitely many pieces; then using the notion of graph retraction, we obtain an easily understandable description of the graphs in each piece.

Disassemble Byte Sequence Using Graph Attention Network

Disassembly is the basis of static analysis of binary code and is used in malicious code detection, vulnerability mining, software optimization, etc. Disassembly of arbitrary suspicious code blocks (e.g., for suspicious traffic packets intercepted by the network) is a difficult task. Traditional disassembly methods require manual specification of the starting address and cannot automate the disassembly of arbitrary code blocks. In this paper, we propose a disassembly method based on code extension selection network by combining traditional linear sweep and recursive traversal methods. First, each byte of a code block is used as the disassembly start address, and all disassembly results (control flow graphs) are combined into a single flow graph. Then a graph attention network is trained to pick the correct subgraph (control flow graph) as the final result. In the experiment, the compiler-generated executable file, as well as the executable file generated by hand-written assembly code, the data file and the byte sequence intercepted by the code segment were tested, and the disassembly accuracy was 93%, which can effectively distinguish the code from the data.

A Batch-dynamic Suitor Algorithm for Approximating Maximum Weighted Matching

Matching is a popular combinatorial optimization problem with numerous applications in both commercial and scientific fields. Computing optimal matchings w.r.t. cardinality or weight can be done in polynomial time; still, this task can become infeasible for very large networks. Thus, several approximation algorithms that trade solution quality for a faster running time have been proposed. For networks that change over time, fully dynamic algorithms that efficiently maintain an approximation of the optimal matching after a graph update have been introduced as well. However, no semi- or fully dynamic algorithm for (approximate) maximum weighted matching has been implemented. In this article, we focus on the problem of maintaining a \( 1/2 \) -approximation of a maximum weighted matching (MWM) in fully dynamic graphs. Limitations of existing algorithms for this problem are (i) high constant factors in their time complexity, (ii) the fact that none of them supports batch updates, and (iii) the lack of a practical implementation, meaning that their actual performance on real-world graphs has not been investigated. We propose and implement a new batch-dynamic \( 1/2 \) -approximation algorithm for MWM based on the Suitor algorithm and its local edge domination strategy [Manne and Halappanavar, IPDPS 2014]. We provide a detailed analysis of our algorithm and prove its approximation guarantee. Despite having a worst-case running time of \( \mathcal {O}(n + m) \) for a single graph update, our extensive experimental evaluation shows that our algorithm is much faster in practice. For example, compared to a static recomputation with sequential Suitor , single-edge updates are handled up to \( 10^5\times \) to \( 10^6\times \) faster, while batches of \( 10^4 \) edge updates are handled up to \( 10^2\times \) to \( 10^3\times \) faster.

Graphical analysis of guideline adherence to detect systemwide anomalies in HIV diagnostic testing.

Analyses of electronic medical databases often compare clinical practice to guideline recommendations. These analyses have a limited ability to simultaneously evaluate many interconnected medical decisions. We aimed to overcome this limitation with an alternative method and apply it to the diagnostic workup of HIV, where misuse can contribute to HIV transmission, delay care, and incur unnecessary costs. We used graph theory to assess patterns of HIV diagnostic testing in a national healthcare system. We modeled the HIV diagnostic testing guidelines as a directed graph. Each node in the graph represented a test, and the edges pointed from one test to the next in chronological order. We then graphed each patient's HIV testing. This set of patient-level graphs was aggregated into a single graph. Finally, we compared the two graphs, the first representing the recommended approach to HIV diagnostic testing and the second representing the observed patterns of HIV testing, to assess for clinical practice deviations. The HIV diagnostic testing of 1.643 million patients provided 8.790 million HIV diagnostic test results for analysis. Significant deviations from recommended practice were found including the use of HIV resistance tests (n = 3,007) and HIV nucleic acid tests (n = 16,567) instead of the recommended HIV screen. We developed a method that modeled a complex medical scenario as a directed graph. When applied to HIV diagnostic testing, we identified deviations in clinical practice from guideline recommendations. The model enabled the identification of intervention targets and prompted systemwide policy changes to enhance HIV detection.

Study-GNN: A Novel Pipeline for Student Performance Prediction Based on Multi-Topology Graph Neural Networks

Student performance prediction has attracted increasing attention in the field of educational data mining, or more broadly, intelligent education or “AI + education”. Accurate performance prediction plays a significant role in solving the problem of a student dropping out, promoting personalized learning and improving teaching efficiency, etc. Traditional student performance prediction methods usually ignore the potential (underlying) relationship among students. In this paper, we use graph structure to reflect the students’ relationships and propose a novel pipeline for student performance prediction based on newly-developed multi-topology graph neural networks (termed MTGNN). In particular, we propose various ways for graph construction based on similarity learning using different distance metrics. Based on the multiple graphs of different topologies, we design an MTGNN module, as a key module in the pipeline, to deal with the semi-supervised node classification problem where each node represents a student (and the node label is the student’s performance, e.g., Pass/Fail/Withdrawal). An attention-based method is developed to produce the unified graph representation in MTGNN. The effectiveness of the proposed pipeline is verified in a case study, where a real-world educational dataset and several existing approaches are used for performance comparison. The experiment results show that, compared with some traditional machine learning methods and the vanilla graph convolutional network with only a single graph topology, our proposed pipeline works effectively and favorably in student performance prediction.

Visualization of aggregated information to support class-level software evolution

Context:Software is inherently prone to constant change, especially in the source code, making it difficult for developers to keep track of changes performed over time and to fully understand their implications. Objective:To this end, we present an Eclipse plug-in for visualizing heterogeneous information, collected from multiple sources, at different levels of granularity. This visualization provides a single graphical representation of a system’s change histories over multiple versions, allowing developers to identify the previous and present dependencies in the system, while adding new or removing and modifying the current functionalities of a software. Summarizing and associating the relevant changes in a single graph, further supports developers, not familiar with the overall system, to conduct a self-study and explore the systems design and changes of its functionality over time. Method:Our tool, DejaVu, initially infers and further visualizes change scenarios that have been applied to a given class in source code, across multiple versions of a software. DejaVu additionally augments the change information with prior commits from GitHub repositories, as well as associated issues from Jira issue tracking system. Evaluation:As part of the evaluation, we conducted a controlled experiment, recruiting participants with research or industrial programming experiences. The participants were asked to investigate and assess a set of change stories with and without the use of the DejaVu. As such, we empirically evaluated the impact of DejaVu in alleviating developers’ understanding of code class-level changes across multiple versions. Results:Our results showed an average of 52% reduction in completion time and a 51% increase in correctness of several change-comprehension tasks once, users adopted DejaVu in comparison to the manual completion of the same tasks. A student’s t-test verified the significant improvement in time and correctness of the tasks with p-values of 0.01 and 0.002. Conclusion:Visualizing aggregated information from multiple sources provides developers with a more comprehensive intuition of the change and its rationale, facilitating software maintenance tasks.

Differentiated graph regularized non-negative matrix factorization for semi-supervised community detection

Semi-supervised non-negative matrix factorization (Semi-NMF) has been widely used in community detection by employing the side information. However, the graph used in previous Semi-NMF methods only takes into account single graph construction, being aware of specific similarity measurements among the community nodes. In this paper, we propose a novel approach, named Differentiated Graph regularized Non-negative Matrix Factorization (DGNMF), for semi-supervised community detection by leveraging the paired constraints between network nodes. In particular, the similarity and dissimilarity constraints on must-link and cannot-link data samples are imposed respectively to construct a differentiated graph that is involved into the proposed algorithm in a form of dual-sparse NMF problem. To solve the optimization problem, we propose an alternate iterative algorithm and demonstrate its convergence theoretically. Extensive experiments on two artificial networks and ten real-world networks show that the proposed DGNMF can effectively improve the accuracy of community detection compared with the state-of-the-art NMF-based approaches.

ConAs-GRNs: Sentiment Classification with Construction-Assisted Multi-Scale Graph Reasoning Networks

Traditional neural networks have limited capabilities in modeling the refined global and contextual semantics of emotional texts and usually ignore the dependencies between different emotional words. To address this limitation, this paper proposes a construction-assisted multi-scale graph reasoning network (ConAs-GRNs), which explores the details of the contextual semantics as well as the emotional dependencies between emotional texts from multiple aspects by focusing on the salient emotional information. In this network, an emotional construction-based multi-scale topological graph is used to describe multiple aspects of emotional dependency, and a sentence dependency tree is utilized to construct a relationship graph based on emotional words and texts. Then, the transfer learning and pooling learning on the topology map is performed. In our case, a weighted edge reduction strategy is used to aggregate the adjacency information which enables the internal transfer of semantic information in a single graph. Moreover, to implement the inter-graph transfer of semantic information, we rely on the construction structure to coordinate the heterogeneous graph information. The extensive experiments conducted on two baseline datasets, SemEval 2014 and ACL-14, demonstrate that the proposed ConAs-GRNs can effectively coordinate and integrate the heterogeneous information from within constructions.

MGLNN: Semi-supervised learning via Multiple Graph Cooperative Learning Neural Networks

In many machine learning applications, data are coming with multiple graphs, which is known as the multiple graph learning problem. The problem of multiple graph learning is to learn consistent representation by exploiting the complementary information of multiple graphs. Graph Learning Neural Networks (GLNNs) have been demonstrated powerfully for graph data representation and semi-supervised classification tasks. However, Existing GLNNs are mainly developed for single graph data which cannot be utilized for multiple graph data representation. In this paper, we propose a novel learning framework, called Multiple Graph Learning Neural Networks (MGLNN), for multiple graph learning and multi-view semi-supervised classification. The goal of MGLNN is to learn an optimal graph structure from multiple graph structures that best serves GNNs’ learning which integrates multiple graph learning and GNNs’ representation simultaneously. The proposed MGLNN is a general framework which can incorporate any specific GNN model to deal with multiple graphs. A general algorithm has also been developed to optimize/train the proposed MGLNN model. Experimental results on several datasets demonstrate that MGLNN outperforms some other related methods on semi-supervised classification tasks.

OWGraMi: Efficient method for mining weighted subgraphs in a single graph

Recently, the problem of mining weighted subgraphs from a weighted single graph has become a vital issue because weighted graphs are generally used to restore, simulate or monitor many complex and large systems in which each object has a different role/level. This field has attracted the attention of numerous researchers, and within related studies Weighted Graph Mining (WeGraMi) can be considered as the state-of-the-art method. However, WeGraMi lacks a strategy to prune unweighted candidate subgraphs early in the process and needs a lot of time to compute the weight for all mined frequent subgraphs. In this paper, we optimize the WeGraMi algorithm with the use of two effective strategies in a new method, which we call Optimized Weighted Graph Mining (OWGraMi). Firstly, we use a strategy to prune all frequent edges which cannot reach the weighting threshold, and with this method we can decrease the number of unweighted candidates. Secondly, we reuse the weight of parent subgraphs when computing the weight for their child subgraphs, as this can reduce the running time for the mining process. On two real graph datasets, directed as well as undirected, our experiments show that both the running time and memory requirements for OWGraMi can be reduced significantly in comparison to the original algorithm.

An efficient and scalable approach for mining subgraphs in a single large graph

An efficient and scalable approach for mining subgraphs in a single large graph

A Biomedical Open Knowledge Network Harnesses the Power of AI to Understand Deep Human Biology

Knowledge representation and reasoning (KR&amp;R) has been successfully implemented in many fields to enable computers to solve complex problems with AI methods. However, its application to biomedicine has been lagging in part due to the daunting complexity of molecular and cellular pathways that govern human physiology and pathology. In this article, we describe concrete uses of Scalable PrecisiOn Medicine Knowledge Engine (SPOKE), an open knowledge network that connects curated information from thirty-seven specialized and human-curated databases into a single property graph, with 3 million nodes and 15 million edges to date. Applications discussed in this article include drug discovery, COVID-19 research and chronic disease diagnosis, and management.

Multi‐graph convolutional clustering network

The relationship between objects can be described from different angles. Although multiple kinds of relationships make the connections between objects complex, they bring in more discriminative information for the clustering tasks. Therefore, how to effectively fuse multiple kinds of relationships becomes a critical problem. In this paper, we propose a novel Multi-graph Convolutional Clustering Network which deeply explores the feature information of nodes and fuses the multiple kinds of relationships between nodes. Unlike most graph convolutional clustering methods that only exploit the single graph or directly fuse multiple graphs into a unified graph before the graph convolution operation, we firstly build multiple parallelled graph convolution layers for each graph to learn diverse data representations, which fully exploits different statistics information between graphs. Then, a designed multi-graph attention module fuses above data representations and considers the importance of each graph. Besides, the proposed model completes the transition from single graph to multiple graphs, which reduces the dependence of the quality of the single graph and enhances the robustness to graphs. Experimental results verify that the proposed multi-graph convolution clustering performs better than the traditional single-graph convolution clustering.