Graph Convolutional Networks (GCNs) have demonstrated remarkable success in various graph-related tasks. However, recent studies show that GCNs are vulnerable to adversarial attacks on graph structures. Therefore, how to defend against such attacks has become a popular research topic. The current common defense methods face two main limitations: (1) From the data perspective, it may lead to suboptimal results since the structural information is ignored when distinguishing the perturbed edges. (2) From the model perspective, the defenders rely on the low-pass filter of the GCN, which is vulnerable during message passing. To overcome these limitations, this paper analyzes the characteristics of perturbed edges, and based on this we propose a robust defense framework,REDE, to generate the adaptiveReliableDefense graph for multi-channel robust GCN. REDE first uses feature similarity and structure difference to discriminate perturbed edges and generates the defense graph by pruning them. Then REDE designs a multi-channel GCN, which can separately capture the information of different edges and high-order neighbors utilizing different frequency components. Leveraging this capability, the defense graph is adaptively updated at each layer, enhancing its reliability and improving prediction accuracy. Extensive experiments on four benchmark datasets demonstrate the enhanced performance and robustness of our proposed REDE over the state-of-the-art defense methods.

Community search is a personalized community discovery problem aimed at finding densely-connected subgraphs containing the query vertex. In particular, the search for com-munities with high-importance vertices has recently received a great deal of attention. However, existing works mainly focus on conventional homogeneous networks where vertices are of the same type, but are not applicable to heterogeneous information networks (HINs) composed of multi-typed vertices and different semantic relations, such as bibliographic networks. In this paper, we study the problem of high-importance community search in HINs. A novel community model is introduced, named heterogeneous significant community (HSC), to unravel the closely connected vertices of the same type with high attribute values through multiple semantic relationships. An HSC not only maximizes the exploration of indirect relationships across entities of the anchor-type but incorporates their significance. To search the HSCs, we first develop online algorithms by exploiting both segmented-based meta-path expansion and significance incrernent. Specially, a solution space reuse strategy based on structural nesting is designed to boost the efficiency. In addition, we further devise a two-level index to support searching HSCs in optimal time, based on which a space-efficient compact index is proposed. Extensive experiments on real-world large-scale HINs demonstrate that our solutions are effective and efficient for searching HSCs, and the index-based algorithms are 2–4 orders of magnitude faster than online algorithms.

Graph Contrastive Learning (GCL) methods benefit from two key properties: alignment and uniformity, which encourage the representation of related objects together while pushing apart different objects. Most GCL methods aim to preserve alignment and uniformity through random graph augmentation strategies and indiscriminately negative sampling. However, their performance is highly sensitive to graph augmentation, which requires cumbersome trial-and-error and expensive domain-specific knowledge as guidance. Besides, these methods perform negative sampling indiscriminately, which inevitably suffers from sampling bias, i.e., negative samples from the same class as the anchor. To remedy these issues, we propose a unified GCL framework towards Alignment-Uniformity Aware Representation learning (AUAR), which can achieve better alignment while improving uniformity without graph augmentation and negative sampling. Specifically, we propose intra- and inter-alignment loss to align the representations of the node with itself and its cluster centroid to maintain label-invariant. Furthermore, we introduce a uniformity loss with theoretical analysis, which pushes the representations of unrelated nodes from different classes apart and tends to provide informative variance from different classes. Extensive experiments demonstrate that our method gains better performance than existing GCL methods in node classification and clustering tasks across three widely-used datasets.

Graph neural networks (GNNs) have achieved impressive performance in various graph-related tasks. However, recent studies have found that GNNs are vulnerable to adversarial attacks. Node injection attacks (NIA) become an emerging scenario of graph adversarial attacks, where the attacks are performed by injecting malicious nodes into the original graph instead of directly modifying it. In this paper, we focus on a more realistic scenario of NIA, where the attacker is only allowed to inject a small number of nodes to degrade the performance of GNNs with very limited information. We analyze the susceptibility of nodes, and based on this we propose a global node injection attack framework, MaxiMal, to maximize malicious information under a strict black-box setting. MaxiMal first introduces a susceptible-reverse influence sampling strategy to select neighbor nodes that are able to spread malicious information widely. Then contrastive loss is introduced to optimize the objective by updating the edges and features of the injected nodes. Extensive experiments on three benchmark datasets demonstrate the superiority of our proposed MaxiMal over the state-of-the-art approaches.

Predicting the future popularity of individual information cascades has attracted much attention in various application fields. It is significantly important for online advertising, viral marketing, rumor detection, and social recommendation. Most approaches target modeling forwarding path or learning important information from discrete static graph. These methods either extract complicated hand-crafted features that rely on domain knowledge and have lower generality, or devote to modeling the arriving intensity function of each message and cannot be optimized for the final popularity. Despite some approaches trying to utilize the underlying structural information in discrete snapshots, they neglect to model the temporal information that implicitly underlying abundant diffusion patterns. Meanwhile, they ignore the inherent dependencies among forwarding behaviors of users. In this paper, we propose a novel learning framework for popularity prediction via modeling temporal dependencies on dynamic evolution process, called TEDDY. Our framework not only models the temporal evolution in a separate snapshot via multiple sequences temporal encoder, but also captures the inherent temporal dependencies among different snapshots. We have conducted extensive experiments on two real-world datasets, i.e. Sina Weibo and American Physical Society. Experimental results demonstrate that our proposed TEDDY significantly improves the prediction accuracy and is superior to the state-of-the-art approaches.

With the rapid development of web technology, social media platforms have become a breeding ground for rumors. These rumors can threaten people’s health, endanger the economy, and affect the stability of a country. In recent years, to mitigate the problem of rumors, computational detection of rumors has been studied, producing some promising early results. However, how to effectively capture the temporal information of retweet dynamics and the structural information of propagation structure is still neglected. In this article, we innovatively propose a novel Multiview Structural-Temporal Learning Framework for Rumor Detection, MiSTR, to jointly learn the temporal features of retweet dynamics, structural features of propagation graph, and the textual features of source tweet. More specifically, we utilize the timestamp encoding, and timestamp level and sequential level attention mechanisms to learn the temporal correlation among individual retweets. We propose two specific methods to learn the overall representation of propagation structure among users from both microscopic and mesoscopic perspectives. Encouraging empirical results on three real large-scale datasets demonstrate the superiority of our proposed method over the state-of-the-art approaches.

Graphs are the universal data structures for representing the relationships between interconnected objects. They are ubiquitous in a variety of disciplines and domains ranging from computer science, social science, economics, medicine, to bioinformatics. In Recent years, extensive studies have been conducted on the graph analysis techniques. One of the most fundamental challenges of analyzing graphs is effectively representing graphs, which largely determines the performance of many follow-up tasks. This workshop aims to provide a forum for industry and academia to discuss the latest progress on graph representation learning and their applications in different fields. We hope more advanced technologies can be proposed or inspired, and also we expect that the direction of graph representation learning can catch much more attention in both academic and industry.

Modeling and predicting the popularity dynamics of individual user generated items on online social networks has important implications in a wide range of areas. The challenge of this problem comes from the inequality of the popularity of content and the numerous complex factors. Existing works mainly focus on exploring relevant factors for prediction and fitting the time series of popularity dynamics into certain class of functions, while ignoring the underlying arrival process of attentions. Also, the exogenous effect of user activity variation on the platform has been neglected. In this paper, we propose a probabilistic model using an influence-based self-excited Hawkes process (ISEHP) to characterize the process through which individual microblogs gain their popularity. This model explicitly captures three ingredients: the intrinsic attractiveness of a microblog with exponential time decay, the user-specific triggering effect of each forwardings based on the endogenous influence among users, and the exogenous effect from the platform. We validate the ISEHP model by applying it on Sina Weibo, the most popular microblogging network in China. Experimental results demonstrate that our proposed model consistently outperforms existing prediction models.

Crowd simulation explores crowd behavior in virtual environments, which has been extensively studied in many areas, such as safety and civil engineering, transportation, social science, and entertainment industry. In this paper, an improved potential field method is proposed to achieve the real-time crowd simulation, which is composed of the global navigation with Dijkstra's algorithm and the potential field based local navigation. First, a region separation is performed to divide the environment into a set of triangles, and thus a topological graph can be built with the triangles as vertices. Then a velocity-density model is introduced for improving the speed controlling mechanism and solving the "maximum speed dilemma" which means the velocity of an individual derived by potential field will be stuck into the maximum due to the ill speed control. Since the movement of an individual in the crowd is influenced by the socio-psychological forces, the individuals' actions express the group attributes. In order to represent the group attributes in the crowd, the repulsive potential function is improved in this paper. Experiments have been carried out and the results show that the improved potential field based method can simulate the crowd in real time and avoid the "maximum speed dilemma".