Query Nodes Research Articles

Conditional Community Search Based on Weight Information

Community search aims to identify cohesive subgraphs containing user-given query nodes in social networks. As information technology develops, user demands for community search have become increasingly sophisticated. The searched communities must not only meet the structural cohesiveness requirements but also adhere to some complex search conditions based on Boolean expressions. For example, certain desired nodes should be contained in the communities, while certain undesired nodes cannot exist in the communities, which is called conditional community search. However, existing solutions for conditional community search often introduce some undesired nodes into the identified communities and exhibit relatively low search efficiency. To overcome these drawbacks, therefore, this paper investigates the problem of conditional community search based on weight information. First, we refine the original problem definition of conditional community search and outline the need for an improved algorithm for calculating the weights of the nodes. Then, we explore two novel algorithms for searching conditional communities based on calculated weight information. Finally, we conduct extensive experiments on several real-world datasets to verify the accuracy and efficiency of our proposed searching algorithms.

Locally-adaptive mapping for network alignment via meta-learning

Network alignment (NA), discovering anchor nodes that represent the same entities across different networks, plays a fundamental role in information fusion. Most existing embedding-based methods rarely study the alignment module, which learns a global mapping to unify embedding spaces. However, global mapping is a holistic solution for nodes, incapable of optimally projecting each node, thus deteriorating alignment accuracy. To solve this problem, this paper proposes a local mapping framework named MANA, which fine-tunes global mapping by meta-learning to obtain node-level local mappings. One advantage of MANA is that it tailors local mapping adapted to the embedded locality of each node while maintaining the general knowledge of global mapping. The main challenge of applying meta-learning to network alignment is paradigm incompatibility; that is, how to construct effective meta-tasks and support sets for zero-shot NA. Therefore, the paper constructs meta-tasks with similarity-based support sets. A support set is taken from pairs of anchor nodes, the source nodes of which are embedded close to the query node. Our framework can be applied to existing mapping-based NA models. Experimental results show that mapping-based models with MANA improve evaluation scores by 1%–59% relative to their original models, demonstrating the effectiveness of local mapping. Some simple mapping-based models improved by MANA even outperform sophisticated sharing-based NA approaches.

Inductive Attributed Community Search: To Learn Communities Across Graphs

Attributed community search (ACS) aims to identify subgraphs satisfying both structure cohesiveness and attribute homogeneity in attributed graphs, for a given query that contains query nodes and query attributes. Previously, algorithmic approaches deal with ACS in a two-stage paradigm, which suffer from structural inflexibility and attribute irrelevance. To overcome this problem, recently, learning-based approaches have been proposed to learn both structures and attributes simultaneously as a one-stage paradigm. However, these approaches train a transductive model which assumes the graph to infer unseen queries is as same as the graph used for training. That limits the generalization and adaptation of these approaches to different heterogeneous graphs. In this paper, we propose a new framework, Inductive Attributed Community Search, IACS , by inductive learning, which can be used to infer new queries for different communities/graphs. Specifically, IACS employs an encoder-decoder neural architecture to handle an ACS task at a time, where a task consists of a graph with only a few queries and corresponding ground-truth. We design a three-phase workflow, "training-adaptation-inference", which learns a shared model to absorb and induce prior effective common knowledge about ACS across different tasks. And the shared model can swiftly adapt to a new task with small number of ground-truth. We conduct substantial experiments in 7 real-world datasets to verify the effectiveness of IACS for CS/ACS. Our approach IACS achieves 28.97% and 25.60% improvements in F1-score on average in CS and ACS, respectively.

Efficient Unsupervised Community Search with Pre-Trained Graph Transformer

Community search has aroused widespread interest in the past decades. Among existing solutions, the learning-based models exhibit outstanding performance in terms of accuracy by leveraging labels to 1) train the model for community score learning, and 2) select the optimal threshold for community identification. However, labeled data are not always available in real-world scenarios. To address this notable limitation of learning-based models, we propose a pre-trained graph Trans former based community search framework that uses Zero label (i.e., unsupervised), termed TransZero. TransZero has two key phases, i.e., the offline pre-training phase and the online search phase. Specifically, in the offline pre-training phase, we design an efficient and effective community search graph transformer ( CSGphormer ) to learn node representation. To pre-train CSGphormer without the usage of labels, we introduce two self-supervised losses, i.e., personalization loss and link loss, motivated by the inherent uniqueness of node and graph topology, respectively. In the online search phase, with the representation learned by the pre-trained CSGphormer , we compute the community score without using labels by measuring the similarity of representations between the query nodes and the nodes in the graph. To free the framework from the usage of a label-based threshold, we define a new function named expected score gain to guide the community identification process. Furthermore, we propose two efficient and effective algorithms for the community identification process that run without the usage of labels. Extensive experiments over 10 public datasets illustrate the superior performance of TransZero regarding both accuracy and efficiency.

Dual-Path Graph Neural Network with Adaptive Auxiliary Module for Link Prediction.

Link prediction, which has important applications in many fields, predicts the possibility of the link between two nodes in a graph. Link prediction based on Graph Neural Network (GNN) obtains node representation and graph structure through GNN, which has attracted a growing amount of attention recently. However, the existing GNN-based link prediction approaches possess some shortcomings. On the one hand, because a graph contains different types of nodes, it leads to a great challenge for aggregating information and learning node representation from its neighbor nodes. On the other hand, the attention mechanism has been an effect instrument for enhancing the link prediction performance. However, the traditional attention mechanism is always monotonic for query nodes, which limits its influence on link prediction. To address these two problems, a Dual-Path Graph Neural Network (DPGNN) for link prediction is proposed in this study. First, we propose a novel Local Random Features Augmentation for Graph Convolution Network as a baseline of one path. Meanwhile, Graph Attention Network version 2 based on dynamic attention mechanism is adopted as a baseline of the other path. And then, we capture more meaningful node representation and more accurate link features by concatenating the information of these two paths. In addition, we propose an adaptive auxiliary module for better balancing the weight of auxiliary tasks, which brings more benefit to link prediction. Finally, extensive experiments verify the effectiveness and superiority of our proposed DPGNN for link prediction.

Enhancing Semi-supervised Domain Adaptation via Effective Target Labeling

Existing semi-supervised domain adaptation (SSDA) models have exhibited impressive performance on the target domain by effectively utilizing few labeled target samples per class (e.g., 3 samples per class). To guarantee an equal number of labeled target samples for each class, however, they require domain experts to manually recognize a considerable amount of the unlabeled target data. Moreover, as the target samples are not equally informative for shaping the decision boundaries of the learning models, it is crucial to select the most informative target samples for labeling, which is, however, impossible for human selectors. As a remedy, we propose an EFfective Target Labeling (EFTL) framework that harnesses active learning and pseudo-labeling strategies to automatically select some informative target samples to annotate. Concretely, we introduce a novel sample query strategy, called non-maximal degree node suppression (NDNS), that iteratively performs maximal degree node query and non-maximal degree node removal to select representative and diverse target samples for labeling. To learn target-specific characteristics, we propose a novel pseudo-labeling strategy that attempts to label low-confidence target samples accurately via clustering consistency (CC), and then inject information of the model uncertainty into our query process. CC enhances the utilization of the annotation budget and increases the number of “labeled” target samples while requiring no additional manual effort. Our proposed EFTL framework can be easily coupled with existing SSDA models, showing significant improvements on three benchmarks

Efficient and effective [formula omitted]-core-based community search over attributed heterogeneous information networks

Given a heterogeneous information network (HIN) G and a query node q, community search (CS) over an HIN identifies a cohesive subgraph from G that contains q. Although HINs with attributes on nodes (called AHINs) are prevalent today, the CS over AHINs (CS-AHIN) is ignored in the literature. Though we can convert an AHIN to an attributed homogeneous graph given a meta-path, then apply the CS approaches for attributed homogeneous graphs to solve CS-AHIN, it is problematic for two reasons. (1) Complete graph conversion is time-consuming and unnecessary, because the search only involves the query node's neighborhood, not the entire graph. (2) Existing attribute cohesiveness metrics are not strict enough to reflect substantial similarities among the community's pairwise nodes. To resolve this, we define the CS-AHIN problem atop a strict attribute cohesiveness metric that supports textual and numerical attributes simultaneously. We show the problem is NP-hard. To address it, we propose an exact baseline to return the global optimal result. Then, we propose three heuristic algorithms using a general greedy search framework to speed up the efficiency. Moreover, we present a cohesiveness-aware proximity graph-based index to boost the performance. Comprehensive experimental studies on various real-world datasets demonstrate our method's superiority.

Neural Attributed Community Search at Billion Scale

Community search has been extensively studied in the past decades. In recent years, there is a growing interest in attributed community search that aims to identify a community based on both the query nodes and query attributes. A set of techniques have been investigated. Though the recent methods based on advanced learning models such as graph neural networks (GNNs) can achieve state-of-the-art performance in terms of accuracy, we notice that 1) they suffer from severe efficiency issues; 2) they directly model community search as a node classification problem and thus cannot make good use of interdependence among different entities in the graph. Motivated by these, in this paper, we propose a new neur AL attr I buted Community s E arch model for large-scale graphs, termed ALICE. ALICE first extracts a candidate subgraph to reduce the search scope and subsequently predicts the community by the Consistency-aware Net, termed ConNet. Specifically, in the extraction phase, we introduce the density sketch modularity that uses a unified form to combine the strengths of two existing powerful modularities, i.e., classical modularity and density modularity. Based on the new modularity metric, we first adaptively obtain the candidate subgraph, formed by the k-hop neighbors of the query nodes, with the maximum modularity. Then, we construct a node-attribute bipartite graph to take attributes into consideration. After that, ConNet adopts a cross-attention encoder to encode the interaction between the query and the graph. The training of the model is guided by the structure-attribute consistency and the local consistency to achieve better performance. Extensive experiments over 11 real-world datasets including one billion-scale graph demonstrate the superiority of ALICE in terms of accuracy, efficiency, and scalability. Notably, ALICE can improve the F1-score by 10.18% on average and is more efficient on large datasets in comparison to the state-of-the-art. ALICE can finish training on the billion-scale graph within a reasonable time whereas state-of-the-art can not.

Edge-Based Minimal k-Core Subgraph Search

In social networks, k-core is commonly used to measure the stability of a network. When a user in a k-core leaves the network, other users may follow the user to leave. Hence, maintaining a key user is important to keep the stability of a network. It is known that an edge between two users models the relationship between the two users. In some scenarios, maintaining a relationship comes at a cost. Therefore, selectively in maintaining the relationships between users is crucial. In this paper, we for the first time conceive the concept of an edge-based minimal k-core model. An edge-based minimal k-core is a k-core with a minimal number of edges. In other words, removing any edge in an edge-based minimal k-core would make it not be a k-core any more. Based on this model, we proposed two problems, namely, an edge-based minimal k-core subgraph search (EMK-SS) and an edge-based minimal k-core subgraph search with a query node q (EMK-q-SS). Given a graph G, an integer k, and a query node (a key user) q, the EMK-q-SS problem is to find all the edge-based minimal k-cores containing the query node q, and the EMK-SS problem is to find all the edge-based minimal k-cores. We also theoretically prove that the two problems are both NP-complete. To deal with the proposed problems, we design two novel algorithms, namely the edge deletion algorithm and edge extension algorithm. Further, a graph partitioning technique is employed to speed up the computation. Comprehensive experiments on synthetic and real networks are conducted to demonstrate the effect and efficiency of our proposed methods.

A Diagnostic Study of Content-Based Image Retrieval Technique for Studying the CT Images of Lung Nodules and Prediction of Lung Cancer as a Biometric Tool

Content Based Medical Image Retrieval (CBMIR) can be defined as a digital image search using the contents of the images. CBMIR plays a very important part in medical applications such as retrieving CT images and more accurately diagnosing aberrant lung tissues in CT images. The Content-Based Medical Image Retrieval (CBMIR) method might aid radiotherapists in examining a patient's CT image in order to retrieve comparable pulmonary nodes more precisely by utilizing query nodes. Intending a particular query node, the CBMIR system searches a large chest CT image database for comparable nodes. The prime aim of this research is to evaluate an end-to-end method for developing a CBIR system for lung cancer diagnosis.

Robust Graph Meta-Learning via Manifold Calibration with Proxy Subgraphs

Graph meta-learning has become a preferable paradigm for graph-based node classification with long-tail distribution, owing to its capability of capturing the intrinsic manifold of support and query nodes. Despite the remarkable success, graph meta-learning suffers from severe performance degradation when training on graph data with structural noise. In this work, we observe that the structural noise may impair the smoothness of the intrinsic manifold supporting the support and query nodes, leading to the poor transferable priori of the meta-learner. To address the issue, we propose a new approach for graph meta-learning that is robust against structural noise, called Proxy subgraph-based Manifold Calibration method (Pro-MC). Concretely, a subgraph generator is designed to generate proxy subgraphs that can calibrate the smoothness of the manifold. The proxy subgraph compromises two types of subgraphs with two biases, thus preventing the manifold from being rugged and straightforward. By doing so, our proposed meta-learner can obtain generalizable and transferable prior knowledge. In addition, we provide a theoretical analysis to illustrate the effectiveness of Pro-MC. Experimental results have demonstrated that our approach can achieve state-of-the-art performance under various structural noises.

Efficient Parallel Dynamic Overset Unstructured Grid for Flow Simulation with Moving Bodies

The overset grid method for simulation of unsteady flow with moving bodies faces several issues, including low assembly efficiency for a large number of bodies, difficult parallel implementation, and the requirement of manual intervention. To address these issues, in this paper, we develop an efficient, automatic, and robust parallel dynamic overset-unstructured-grid assembly method. It is achieved by parallelizing its two main steps: hole cutting, and identification of interpolation stencils. The hole cutting is simply implemented in parallel by identifying the active zone of each grid with a criterion of the global minimum wall distance. The global minimum wall distance is efficiently calculated on each process with the oriented-bounding-box-based -dimensional (-D) trees of the wall surface grids. For identification of the interpolation stencil, we use the results of the first step, optimally define the interpolation boundary nodes, and determine their candidate donor partitions in each process. This efficiently minimizes the number of query nodes and their donor cell candidates. The donor cell search is efficiently performed in parallel in each candidate donor partition by using the oriented-bounding-box-based -D tree of the field volume grid. Several cases are adopted to test the efficiency and capability of the proposed parallel dynamic overset-unstructured-grid assembly approach.

InfoMax Classification-Enhanced Learnable Network for Few-Shot Node Classification

Graph neural networks have a wide range of applications, such as citation networks, social networks, and knowledge graphs. Among various graph analyses, node classification has garnered much attention. While many of the recent network embedding models achieve promising performance, they usually require sufficient labeled nodes for training, which does not meet the reality that only a few labeled nodes are available in novel classes. While few-shot learning is commonly employed in the vision and language domains to address the problem of insufficient training samples, there are still two characteristics of the few-shot node classification problem in the non-Euclidean domain that require investigation: (1) how to extract the most informative knowledge for a class and use it on testing data and (2) how to thoroughly explore the limited number of support sets and maximize the amount of information transferred to the query set. We propose an InfoMax Classification-Enhanced Learnable Network (ICELN) to address these issues, motivated by Deep Graph InfoMax (DGI), which adapts the InfoMax principle to the summary representation of a graph and the patch representation of a node. By increasing the amount of information that is shared between the query nodes and the class representation, an ICELN can transfer the maximum amount of information to unlabeled data and enhance the graph representation potential. The whole model is trained using an episodic method, which simulates the actual testing environment to ensure the meta-knowledge learned from previous experience may be used for entirely new classes that have not been studied before. Extensive experiments are conducted on five real-world datasets to demonstrate the advantages of an ICELN over the existing few-shot node classification methods.

RelationTrack: Relation-Aware Multiple Object Tracking With Decoupled Representation

Existing online multiple object tracking (MOT) algorithms often consist of two subtasks, detection and re-identification (ReID). In order to enhance the inference speed and reduce the complexity, current methods commonly integrate these double subtasks into a unified framework. Nevertheless, detection and ReID demand diverse features. This issue results in an optimization contradiction during the training procedure. With the target of alleviating this contradiction, we devise a module named Global Context Disentangling (GCD) that decouples the learned representation into detection-specific and ReID-specific embeddings. As such, this module provides an implicit manner to balance the different requirements of these two subtasks. Moreover, we observe that preceding MOT methods typically leverage local information to associate the detected targets and neglect to consider the global semantic relation. To resolve this limitation, we develop a module, referred to as Guided Transformer Encoder (GTE), by combining the powerful reasoning ability of Transformer encoder and deformable attention. Unlike previous works, GTE avoids analyzing all the pixels and only attends to capture the relation between query nodes and a few self-adaptively selected key samples. Therefore, it is computationally efficient. Extensive experiments have been conducted on the MOT16, MOT17 and MOT20 benchmarks to demonstrate the superiority of the proposed MOT framework, namely RelationTrack. The experimental results indicate that RelationTrack has surpassed preceding methods significantly and established a new state-of-the-art performance, e.g., IDF1 of 70.5% and MOTA of 67.2% on MOT20.

Spatial–Temporal Dynamic Graph Attention Network for Skeleton-Based Action Recognition

Human body skeleton, acting as a spatiotemporal graph, is increasing attentions of researchers to adopt graph convolutional networks (GCN) to mine the discriminative features from skeleton joints. However, one of GCN’s flaws is its inability to handle long-distance reliance between joints. In this regard, graph attention network (GAT) was recently suggested, which combines graph convolutions with a self-attention mechanism to extract the most informative joint of a human skeleton and increase the model accuracy. However, GAT can compute only static attention: for each query node, the attention rank is same which severely hurts and limits the expressivity of an attention mechanism. In this work, we present a spatial-temporal dynamic graph attention network (ST-DGAT) to learn the spatial-temporal patterns of skeleton sequences. For dynamic graph attention, we tweak the order of weighted vector operations in GAT, our approach achieves a global approximate attention function, making it strictly superior to GAT. Experiments show that by fixing the order of internal operation of GAT the proposed model achieved better action classification results while maintaining the same computing cost as GAT. The proposed framework has been evaluated on well-known publicly available large-scale datasets NTU60, NTU120, and Kinetics-400, which notably outperforms state-of-the-art (SOTA) results with an accuracy of 96.4%, 88.2%, and 61.0%, respectively.

Adaptive target community search with sample expansion

Target community search aims to search cohesive communities consistent with the user’s preference revealed by query nodes, which is a query-dependent variant of community detection in graph analysis. Most of the existing work either directly expands the target community with user-provided nodes or treats all dimensions of the attribute space as equally important, whereas we argue it is more reasonable to locate high-quality nodes and automatically deduce users’ interests with variation of the relevance for different dimensions of the target community. Towards this end, in this paper, we propose a framework named Adaptive Target Community Search with Sample Expansion (ATCS-SE) for attributed graph, which is able to incorporate user’s preference into community search, thus steering the algorithm to detect more interesting and densely connected communities with high attribute semantic similarity. The algorithm consists of three phases: (i) Augmentation of given prior information: the candidate node path containing a finite sequence of nodes similar to the user-provided query node is established to make up a community backbone; (ii) User preference learning and double view weighting: an adaptive two-level weighting mechanism is designed for community backbone to simultaneously compute weights for both view and variables; and (iii) Refinement of the target community score: the quality score of target communities is defined based on the combination of internal consistency and external separability. Extensive experiments are conducted to show the effectiveness and applied value of the proposed method.

Data structures for categorical path counting queries

Consider an ordinal tree T on n nodes, each of which is assigned a category from an alphabet [σ]={1,2,…,σ}. We preprocess the tree T in order to support categorical path counting queries, which ask for the number of distinct categories occurring on the path in T between two query nodes x and y. For this problem, we propose a linear-space data structure with query time O(nlg⁡lg⁡σlg⁡w), where w=Ω(lg⁡n) is the word size in the word-RAM. As shown in our proof, from the assumption that matrix multiplication cannot be solved in time polynomially faster than cubic (with only combinatorial methods), our result is optimal, save for polylogarithmic speed-ups. For a trade-off parameter 1≤t≤n, we propose an O(n+n2t2)-word, O(tlg⁡lg⁡σlg⁡w) query time data structure. We also consider c-approximate categorical path counting queries, which must return an approximation to the number of distinct categories occurring on the query path, by counting each such category at least once and at most c times. We describe a linear-space data structure that supports 2-approximate categorical path counting queries in O(lg⁡n/lg⁡lg⁡n) time.Next, we generalize the categorical path counting queries to weighted trees. Here, a query specifies two nodes x,y and an orthogonal range Q. The answer to thus formed categorical path range counting query is the number of distinct categories occurring on the path from x to y, if only the nodes with weights falling inside Q are considered. We propose an O(nlg⁡lg⁡n+(n/t)4)-word data structure with O(tlg⁡lg⁡n) query time, or an O(n+(n/t)4)-word data structure with O(tlgϵ⁡n) query time. For an appropriate choice of the trade-off parameter t, this implies a linear-space data structure with O(n3/4lgϵ⁡n) query time. We then extend the approach to the trees weighted with vectors from [n]d, where d is a constant integer greater than or equal to 2. We present a data structure with O(nlgd−1+ϵ⁡n+(n/t)2d+2) words of space and O(tlgd−1⁡n(lg⁡lg⁡n)d−2) query time. For an O(n⋅polylogn)-space solution, one thus has O(n2d+12d+2⋅polylogn) query time.The inherent difficulty revealed by the lower bound we proved motivated us to consider data structures based on sketching. In unweighted trees, we propose a sketching data structure to solve the approximate categorical path counting problem which asks for a (1±ϵ)-approximation (i.e. within 1±ϵ of the true answer) of the number of distinct categories on the given path, with probability 1−δ, where 0<ϵ,δ<1 are constants. The data structure occupies O(n+ntlg⁡n) words of space, for the query time of O(tlg⁡n). For trees weighted with d-dimensional weight vectors (d≥1), we propose a data structure with O((n+ntlg⁡n)lgd⁡n) words of space and O(tlgd+1⁡n) query time.All these problems generalize the corresponding categorical range counting problems in Euclidean space Rd+1, for respective d, by replacing one of the dimensions with a tree topology.

Edge-Aware Graph Neural Network for Multi-Hop Path Reasoning over Knowledge Base.

Multi-hop path reasoning over knowledge base aims at finding answer entities for an input question by walking along a path of triples from graph structure data, which is a crucial branch in the knowledge base question answering (KBQA) research field. Previous studies rely on deep neural networks to simulate the way humans solve multi-hop questions, which do not consider the latent relation information contained in connected edges, and lack of measuring the correlation between specific relations and the input question. To address these challenges, we propose an edge-aware graph neural network for multi-hop path reasoning task. First, a query node is directly added to the candidate subgraph retrieved from the knowledge base, which constructs what we term a query graph. This graph construction strategy makes it possible to enhance the information flow between the question and the nodes for the subsequent message passing steps. Second, question-related information contained in the relations is added to the entity node representations during graph updating; meanwhile, the relation representations are updated. Finally, the attention mechanism is used to weight the contribution from neighbor nodes so that only the information of neighbor nodes related to the query can be injected into new node representations. Experimental results on MetaQA and PathQuestion-Large (PQL) benchmarks demonstrate that the proposed model achieves higher Hit@1 and F1 scores than the baseline methods by a large margin. Moreover, ablation studies show that both the graph construction and the graph update algorithm contribute to performance improvement.

HoSIM: Higher-order Structural Importance based method for multiple local community detection

Local community detection has recently attracted much research attention. Many methods have been proposed for the single local community detection that finds a community containing the given set of query nodes. However, nodes may belong to several communities in the network, and detecting all the communities for the query node set, termed multiple local community detection (MLCD), is more important as it could uncover more potential information. MLCD is also more challenging because when a query node belongs to multiple communities, it is always located in the complicated overlapping region and the marginal region of communities. Accordingly, detecting multiple communities for such nodes by applying seed expansion methods from the nodes is insufficient.This work addresses the MLCD based on higher-order structural importance (HoSI). First, to effectively estimate the influence of higher-order structures, we propose a new variant of random walk called Active Random Walk to measure the HoSI score between nodes. Then, we propose two new metrics to evaluate the HoSI score of a subgraph to a node and the HoSI score of a node, respectively. Based on the proposed metrics, we present a novel algorithm called HoSIM to detect multiple local communities for a query node. HoSIM enforces three-stage processing, namely subgraph sampling, core member identification, and local community detection. The key idea is to utilize HoSI to find and identify the core members of communities relevant to the query node and optimize the generated communities. Extensive experiments illustrate the effectiveness of HoSIM.

Attributed community search based on seed replacement and joint random walk

Community search enables personalized community discovery and has wide applications in real-life scenarios. Existing attributed community search algorithms use personalized information provided by attributes to locate desired community. Though achieved promising results, existing works suffer from two major limitations: (i) the precision of the algorithm decreases significantly when the seed comes from the boundary regions of the community. (ii) Most attributed community search methods mainly take the attribute information as edge weights to reveal semantic strength (e.g., attribute similarity, attribute distance, etc.), but largely ignore that attribute may serve as heterogeneous vertex. To make up for these deficiencies, in this paper, we propose a novel two-stage attributed community search method with seed replacement and joint random walk (SRRW). Specifically, in the seed replacement stage, we replace the initial query node with a core node; in the random walk stage, attributes are taken as heterogeneous nodes and the augmented graph is modeled based on the affiliation of the attributes via an overlapping clustering algorithm. And finally, a joint random walk is performed on the augmented graph to explore the desired local community. We conduct extensive experiments on both synthetic and real-world benchmarks, demonstrating its effectiveness for attributed community search.