Search

Circular RNA (CircRNA)–microRNA (miRNA) interaction (CMI) is an important model for the regulation of biological processes by non-coding RNA (ncRNA), which provides a new perspective for the study of human complex diseases. However, the existing CMI prediction models mainly rely on the nearest neighbor structure in the biological network, ignoring the molecular network topology, so it is difficult to improve the prediction performance. In this paper, we proposed a new CMI prediction method, BEROLECMI, which uses molecular sequence attributes, molecular self-similarity, and biological network topology to define the specific role feature representation for molecules to infer the new CMI. BEROLECMI effectively makes up for the lack of network topology in the CMI prediction model and achieves the highest prediction performance in three commonly used data sets. In the case study, 14 of the 15 pairs of unknown CMIs were correctly predicted. [ABSTRACT FROM AUTHOR]

Machine learning approaches have become a crucial tool in graph analysis. Despite the accurate results of the existing approaches, most of them are not scalable enough to be used in real-world problems. Networks provide two different kinds of information, nodes contents and nodes relations (network structure). Training deep graph neural networks (GNN) over large-scale graphs is challenging due to the limitation of the message passing framework. Graph Convolutional Networks (GCN) work on all node neighbours at once. Furthermore, it is usual to transform node features with a deep neural network before the GC operation. Therefore, the deep transform operation may apply up to hundreds of times for each target node which is heavy computation and hard to batch. This paper presents an abstract framework with two embedding components, the first component embeds node relations, and the second one embeds node contents. The model makes predictions by aggregating these embeddings through a combination component. The presented approach limits the deep transform only to the target node and uses random walk-based embedding instead of the GC operator to reduce the cost. The main goal of the proposed approach is to provide a light framework for the task. To this aim, node relations are embedded based on node neighbourhood structure by a biased variant of the DeepWalk model, called GuidedWalk, and an autoencoder embeds node contents. The experimental results on three well-known datasets show the superiority of the proposed model compared to the state-of-the-art GraphSAGE and TADW models with less computational complexity. On the Citeseer, Cora, and PubMed datasets, the model has achieved 3.23%, 0.88%, and 7.63% improvement in Macro-F1 and 3.25%, 0.7%, and 6.34% improvement in Micro-F1, respectively. Although GNNs are state-of-the-art models, considering node content is their main advantage. This paper shows that even a simple integration of node content to available random walk-based methods improves their performance up to GCNs without increasing the complexity. [ABSTRACT FROM AUTHOR]

With the rapid development of edge computing, online social network services are growing explosively. Social influence plays a critical role in the propagation of social network information. Influence maximization is a key issue in social network analysis. Temporal-aware influence maximization has been emerging in recent years. It integrates temporal information into the classical influence maximization. The purpose for temporal-aware influence maximization is to identify the optimal users in a network that can influence the most individuals under a certain time constraint. Although each node activation is treated equally within a finite time window in existing studies. In reality, the time required for each user to get influenced varies. In contrast, because temporal-aware influence maximization is NP-hard, a number of studies have proposed approximate algorithms based on sample instances to estimate the expected influence spread of nodes. However, these methods require significant computational overheads. Therefore, this study presents learning-based time-decaying influence maximization. Specifically, this study first presents the time-decaying influence maximization, where each user has a utility score that monotonically decreases over time. Subsequently, it develops AttenLSTM-GCN to generate node embeddings that capture structural-temporal information. Then, it adopts Q-learning algorithm to predict the time-decaying influence of nodes. Based on these algorithms, the solution of time-decaying influence maximization can be obtained. Experimental studies on real world networks demonstrate that the proposed model achieves more significant accelerations and high-quality solutions than advanced algorithms. [ABSTRACT FROM AUTHOR]

In recent years, to address the issue of networked data sparsity in node classification tasks, cross-network node classification (CNNC) leverages the richer information from a source network to enhance the performance of node classification in the target network, which typically has sparser information. However, in real-world applications, labeled nodes may be collected from multiple sources with multiple modalities (e.g., text, vision, and video). Naive application of single-source and single-modal CNNC methods may result in sub-optimal solutions. To this end, in this article, we propose a model called Multi-source and Multi-modal Cross-network Deep Network Embedding (M2CDNE) for cross-network node classification. In M2CDNE, we propose a deep multi-modal network embedding approach that combines the extracted deep multi-modal features to make the node vector representations network invariant. In addition, we apply dynamic adversarial adaptation to assess the significance of marginal and conditional probability distributions between each source and target network to make node vector representations label discriminative. Furthermore, we devise to classify nodes in the target network through the related source classifier and aggregate different predictions utilizing respective network weights, corresponding to the discrepancy between each source and target network. Extensive experiments performed on real-world datasets demonstrate that the proposed M2CDNE significantly outperforms the state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

We study the problem of classifying different cooking styles, based on the recipe. The difficulty is that the same food ingredients, seasoning, and the very similar instructions result in different flavors, with different cooking styles. Existing methods have limitations: they mainly focus on homogeneous data (e.g., instruction or image), ignoring heterogeneous data (e.g., flavor compound or ingredient), which certainly hurts the classification performance. This is because collecting enough available heterogeneous data of a recipe is a non-trivial task. In this paper, we present a new heterogeneous data augmentation method to improve classification performance. Specifically, we first construct a heterogeneous recipe graph network to represent heterogeneous data, which includes four main-stream types of heterogeneous data: ingredient, flavor compound, image, and instruction. Then, we draw a sequence of augmented graphs for Semi-Supervised learning through multinomial sampling. The probability distribution of sampling depends on the Cosine distance between the nodes of graph. In this way, we name our approach as Multinomial Sampling Graph Data Augmentation (MS-GDA). Extensive experiments demonstrate that MS-GDA significantly outperforms SOTA baselines on cuisine classification and region prediction with the recipe benchmark dataset. Code is available at https://github.com/LiangzheChen/MS-GDA. [ABSTRACT FROM AUTHOR]

Abstract Circular RNA (CircRNA)–microRNA (miRNA) interaction (CMI) is an important model for the regulation of biological processes by non-coding RNA (ncRNA), which provides a new perspective for the study of human complex diseases. However, the existing CMI prediction models mainly rely on the nearest neighbor structure in the biological network, ignoring the molecular network topology, so it is difficult to improve the prediction performance. In this paper, we proposed a new CMI prediction method, BEROLECMI, which uses molecular sequence attributes, molecular self-similarity, and biological network topology to define the specific role feature representation for molecules to infer the new CMI. BEROLECMI effectively makes up for the lack of network topology in the CMI prediction model and achieves the highest prediction performance in three commonly used data sets. In the case study, 14 of the 15 pairs of unknown CMIs were correctly predicted.

Objective: To elucidate the biological basis of the heart qi deficiency (HQD) pattern, an in-depth understanding of which is essential for improving clinical herbal therapy. Methods: We predicted and characterized HQD pattern genes using the new strategy, TCM-HIN2Vec, which involves heterogeneous network embedding and transcriptomic experiments. First, a heterogeneous network of traditional Chinese medicine (TCM) patterns was constructed using public databases. Next, we predicted HQD pattern genes using a heterogeneous network-embedding algorithm. We then analyzed the functional characteristics of HQD pattern genes using gene enrichment analysis and examined gene expression levels using RNA-seq. Finally, we identified TCM herbs that demonstrated enriched interactions with HQD pattern genes via herbal enrichment analysis. Results: Our TCM-HIN2Vec strategy revealed that candidate genes associated with HQD pattern were significantly enriched in energy metabolism, signal transduction pathways, and immune processes. Moreover, we found that these candidate genes were significantly differentially expressed in the transcriptional profile of mice model with heart failure with a qi deficiency pattern. Furthermore, herbal enrichment analysis identified TCM herbs that demonstrated enriched interactions with the top 10 candidate genes and could potentially serve as drug candidates for treating HQD. Conclusion: Our results suggested that TCM-HIN2Vec is capable of not only accurately identifying HQD pattern genes, but also deciphering the basis of HQD pattern. Furthermore our finding indicated that TCM-HIN2Vec may be further expanded to develop other patterns, leading to a new approach aimed at elucidating general TCM patterns and developing precision medicine.

Abstract Networks provide an understandable and, in the case of small size, visualizable representation of data, which allows us to obtain essential information about the relationships between pairs of nodes, e.g., their distances. In visualization, networks have an alternative two-dimensional vector representation to which various machine-learning methods can be applied. More generally, networks can be transformed into a low-dimensional space using so-called embedding methods, which bridge the gap between network analysis and traditional machine learning by creating numerical representations that capture the essence of the network structure. In this article, we present a new embedding method that uses non-symmetric dependency to find the distance between nodes and applies an iterative procedure to find a satisfactory distribution of nodes in space. For dimension 2 and the visualization of the result, we demonstrate the method’s effectiveness on small networks. For higher dimensions and several larger networks, we present the results of two experiments comparing our results with two well-established methods in the research community, namely node2vec and DeepWalk. The first experiment focuses on a qualitative comparison of the methods, while the second focuses on applying and comparing the classification results to embeddings in a higher dimension. Although the presented method does not outperform the two chosen methods, its results are still comparable. Therefore, we also explain the limitations of our method and a possible way to overcome them.

Heterogeneous graph convolutional networks have gained great popularity in tackling various network analytical tasks on heterogeneous graph data, ranging from link prediction to node classification. However, most existing works ignore the relation heterogeneity with multiplex networks between multi-typed nodes and the different importance of relations in meta-paths for node embedding, which can hardly capture the heterogeneous structure signals across different relations. To tackle this challenge, this work proposes a Multiplex Heterogeneous Graph Convolutional Network (MHGCN+) for multiplex heterogeneous network embedding. Our MHGCN+ can automatically learn the useful heterogeneous meta-path interactions of different lengths with different importance in multiplex heterogeneous networks through multi-layer convolution aggregation. Additionally, we effectively integrate both multi-relation structural signals and attribute semantics into the learned node embeddings with both unsupervised and semi-supervised learning paradigms. Extensive experiments on seven real-world datasets with various network analytical tasks demonstrate the significant superiority of MHGCN+ against state-of-the-art embedding baselines in terms of all evaluation metrics. The source code of our method is available at: https://github.com/FuChF/MHGCN-plus. [ABSTRACT FROM AUTHOR]

In this paper, we address the problem of unsuperised social network embedding, which aims to embed network nodes, including node attributes, into a latent low dimensional space. In recent methods, the fusion mechanism of node attributes and network structure has been proposed for the problem and achieved impressive prediction performance. However, the non-linear property of node attributes and network structure is not efficiently fused in existing methods, which is potentially helpful in learning a better network embedding. To this end, in this paper, we propose a novel model called ASM (Adaptive Specific Mapping) based on encoder-decoder framework. In encoder, we use the kernel mapping to capture the non-linear property of both node attributes and network structure. In particular, we adopt two feature mapping functions, namely an untrainable function for node attributes and a trainable function for network structure. By the mapping functions, we obtain the low dimensional feature vectors for node attributes and network structure, respectively. Then, we design an attention layer to combine the learning of both feature vectors and adaptively learn the node embedding. In encoder, we adopt the component of reconstruction for the training process of learning node attributes and network structure. We conducted a set of experiments on seven real-world social network datasets. The experimental results verify the effectiveness and efficiency of our method in comparison with state-of-the-art baselines. [ABSTRACT FROM AUTHOR]

Academic collaboration can break through the geographical limitations of scholars and promote academic output among scholars. Academic big data will provide an important data source for more comprehensive and accurate modeling scholars due to the coexistence environment of various academic entities. Based on academic big data, an end-to-end model HCSP was proposed for predicting collaborative relationships among scholars. HCSP models scholars from two aspects: content-based features and structure-based features, and calculate the similarity between scholars based on this to predict whether there will be academic collaboration between scholars. When learning the content-based features of scholars, HCSP utilizes LSTM and multi-head attention mechanism to extract the overall and recent research interests of scholars, to capture the diversity of scholars' research interests. When learning the structure-based features of scholars, HCSP adopts a subgraph sampling strategy based on meta paths to model the structural features of scholar nodes in heterogeneous academic network. By integrating scholars' content-based and structure-based features, HCSP calculates the similarity between scholars to determine whether there will be a collaborative relationship between them. The experimental results indicate that the HCSP model achieves better prediction performance compared to the baseline models. It can be seen that integrating scholars' content-based and structure-based characteristics can indeed provide a richer and more effective modeling basis for predicting their academic collaborative relationships. [ABSTRACT FROM AUTHOR]

Studies have indicated that focusing solely on pairwise interactions between two nodes disregards the associativity among multi-nodes in the network’s local structure. This associativity can be seen as dependencies among nodes, where certain edges’ presence depends on the path leading to it. Examinations on diverse datasets have approved that the variable order of chained dependencies allows for the preservation of structure information, which enables the reconstruction of the original network into a Higher-Order Network (HON) with improved quality of network representation. This paper proposes a Density-based Higher-Order Network Embedding (DHONE) algorithm, which integrates the concept of higher-order density into the network-embedding process in order to classify the contribution of different orders of dependencies. Through the construction of a novel and effective higher-order adjacency matrix, DHONE steadily improves the accuracy of network representation learning. Experimental results demonstrate DHONEs proficiency in improving embedding accuracy and overall algorithm robustness. Furthermore, grounded in the concept of higher-order density proposed herein, numerous dependencies have been discerned within the network generated from trajectories, potentially indicating the role of multi-node structures in networks. [ABSTRACT FROM AUTHOR]

Dynamic network captures time-varying interactions among multiple entities at different time points, and detecting its structural change points is of central interest. This article proposes a novel method for detecting change points in dynamic networks by fully exploiting the latent network structure. The proposed method builds upon a tensor-based embedding model, which models the time-varying network heterogeneity through an embedding matrix. A fused lasso penalty is equipped with the tensor decomposition formulation to estimate the embedding matrix and a power update algorithm is developed to tackle the resultant optimization task. The error bound of the obtained estimated embedding matrices is established without incurring the computational-statistical gap. The proposed method also produces a set of estimated change points, which, coupled with a simple screening procedure, assures asymptotic consistency in change point detection under much milder assumptions. Various numerical experiments on both synthetic and real datasets also support its advantage. for this article are available online. [ABSTRACT FROM AUTHOR]

Understanding Protein - Protein interaction networks, which show the interactions between proteins involved in tasks that are very important for our organisms such as structural support, storage, signal transduction and defence, provides a better understanding of cellular processes. One of the important studies carried out for this purpose is to try to detect protein complexes from protein - protein interaction networks. Supervised and unsupervised machine learning methods were used to detect protein complexes. It is known that the machine learning methods used produce better performance when more than one method is used together. Based on this knowledge, a method that detects protein complexes from protein-protein interaction networks is proposed in this study. The method first weights protein-protein interaction networks using biological and topological properties of proteins. Then it estimates local and global protein complex core. Then it builds a protein complex detection model using the structural modularity of proteins and the voting regression model. We predict that XGB regression, gaussian process regression, catboost regression and histogram-based gradient boosting regression supervised learning methods can achieve more successful results when used together in the voting regression model. When we compare the success of the model with other models, it has shown the best performance many times among the compared models. [ABSTRACT FROM AUTHOR]

Network embedding has been extensively used in several practical applications and achieved great success. However, existing studies mainly focus on single task or single view and cannot obtain deeper relevant information for accomplishing tasks. In this paper, a novel approach is proposed to address the problem of insufficient information consideration in network embedding, which is termed multi-task-oriented adaptive dual-channel graph convolutional network (TAD-GCN). We firstly use kNN graph construction method to generate three views for each network dataset. Then, the proposed TAD-GCN contains dual-channel GCN which can extract the specific and shared embeddings from multiple views simultaneously, and attention mechanism is adopted to fuse them adaptively. In addition, we design similarity constraint and difference constraint to further enhance their semantic similarity and ensure that they capture the different information. Lastly, a multi-task learning module is introduced to solve multiple tasks simultaneously and optimize the model with its losses. The experimental results demonstrate that our model TAD-GCN not only completes multiple downstream tasks at the same time, but also achieves excellent performance compared with eight state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Ambidextrous innovation synergy is an effective way for new entrants and R&D entities to break the blockade of key core technologies. This paper constructs a tripartite evolutionary game model of new entrants, the R&D entity, and monopoly enterprises under the monopoly situation of key core technologies, discusses the dynamic equilibrium process of how new entrants cooperate with the R&D entity to carry out the ambidextrous innovation synergy strategy, and extends the model to the policy subsidy situations of different development stages of key core technology. The results show that the monopoly of key core technologies enhances the original innovation search ability of new entrants and promotes the evolution of enterprise imitation innovation to the exploratory innovation strategy. In the basic research stage of key core technology, the exploratory innovation strategy of new entrants is more sensitive to the cost of network embedding and the original innovation knowledge search. New entrants prefer the imitation innovation strategy, and policy subsidies have no significant effect on exploratory innovation. In the promotion stage of the key core technology market, fiscal and tax subsidies can more easily promote the evolution of new entrants from the imitative innovation strategy to the exploratory innovation strategy than R&D subsidies, and network embeddedness can induce enterprises to carry out exploratory innovation only when a certain threshold is reached. In addition, this paper discusses the influence mechanism of monopoly enterprises' suppression intensities and key core technology breakthrough probabilities on the evolution equilibrium of new entrants' ambidextrous innovation synergy strategies.

In power systems, failures of vulnerable lines can trigger large-scale cascading failures, and vulnerability assessment is dedicated to locating these lines and reducing the risks of such failures. Based on a structure and attribute network embedding (SANE) algorithm, a novel quantitative vulnerability analysis method is proposed to identify vulnerable lines in this research. First, a two-layered random walk network with topological and electrical properties of transmission lines is established. Subsequently, based on the weighted degree of nodes in the two-layered network, the inter-layer and intra-layer walking transition probabilities are developed to obtain walk sequences. Then, a Word2Vec algorithm is applied to obtain low-dimension vectors representing transmission lines, according to obtained walk sequences for calculating the vulnerability index of transmissions lines. Finally, the proposed method is compared with three widely used methods in two test systems. Results show the network embedding based method is superior to those comparison methods and can provide guidance for identifying vulnerable lines.

Heterogeneous Information Networks have bunches of rich secret information that assist us in the creation of successful recommendation frameworks. A Heterogeneous Information Network (HIN) includes useful knowledge required for a recommendation system, and the network embedding is the common strategy for getting useful knowledge out of a HIN to be used in recommendation platforms. Although user and business nodes have been used in HINs, review contents have not been used. In this work, we use review nodes in HINs in addition to user and business nodes. Since written reviews provide valuable insights into points of interest within recommendation systems, integrating review nodes into HINs allows us to assess their impact on recommendation systems. Specifying meaningful meta-paths aids in extracting hidden information within a heterogeneous information network. While user and business nodes are typically utilized for specifying meaningful meta-paths, review nodes have been overlooked. We introduce new meta-paths incorporating review nodes to uncover hidden information in heterogeneous information networks. These meta-paths are leveraged to enhance the recommendation system’s performance. This study endeavors to amalgamate rich written reviews with heterogeneous information networks and analyze their effects on recommendation systems. Our experiments demonstrate that incorporating review texts improves the recommendation system, particularly when selecting meaningful meta-paths. Augmenting HINs with reviews facilitates the capture of additional relational information between users and businesses, thereby enhancing the recommendation model. This underscores the benefits of consolidating interaction information within HIN features for superior recommendation outcomes.

The high-throughput genomic and proteomic scanning approaches allow investigators to measure the quantification of genome-wide genes (or gene products) for certain disease conditions, which plays an essential role in promoting the discovery of disease mechanisms. The high-throughput approaches often generate a large gene list of interest (GOIs), such as differentially expressed genes/proteins. However, researchers have to perform manual triage and validation to explore the most promising, biologically plausible linkages between the known disease genes and GOIs (disease signals) for further study. Here, to address this challenge, we proposed a network-based strategy DDK-Linker to facilitate the exploration of disease signals hidden in omics data by linking GOIs to disease knowns genes. Specifically, it reconstructed gene distances in the protein–protein interaction (PPI) network through six network methods (random walk with restart, Deepwalk, Node2Vec, LINE, HOPE, Laplacian) to discover disease signals in omics data that have shorter distances to disease genes. Furthermore, benefiting from the establishment of knowledge base we established, the abundant bioinformatics annotations were provided for each candidate disease signal. To assist in omics data interpretation and facilitate the usage, we have developed this strategy into an application that users can access through a website or download the R package. We believe DDK-Linker will accelerate the exploring of disease genes and drug targets in a variety of omics data, such as genomics, transcriptomics and proteomics data, and provide clues for complex disease mechanism and pharmacological research. DDK-Linker is freely accessible at http://ddklinker.ncpsb.org.cn/. [ABSTRACT FROM AUTHOR]

Due to the ubiquity of graph-structured data, Graph Neural Network (GNN) have been widely used in different tasks and domains and good results have been achieved in tasks such as node classification and link prediction. However, there are still many challenges in representation learning of heterogeneous networks. Existing graph neural network models are partly based on homogeneous graphs, which do not take into account the rich semantic information of nodes and edges due to their different types; And partly based on heterogeneous graphs, which require predefined meta-structures (include meta-paths and meta-graphs) and do not take into account the different effects of different meta-structures on node representation. In this paper, we propose the MS-GAN model, which consists of four parts: graph structure learner, graph structure expander, graph structure filter and graph structure parser. The graph structure learner automatically generates a graph structure consisting of useful meta-paths by selecting and combining the sub-adjacent matrices in the original graph using a 1 × 1 convolution. The graph structure expander further generates a graph structure containing meta-graphs by Hadamard product based on the previous step. The graph structure filterer filters out graph structures that are more effective for downstream classification tasks based on diversity. The graph structure parser assigns different weights to graph structures consisting of different meta-structures by a semantic hierarchical attention. Finally, through experiments on four datasets and meta-structure visualization analysis, it is shown that MS-GAN can automatically generate useful meta-structures and assign different weights to different meta-structures. • Semi-supervised learning based graph neural network. • Automatically generate meta-structures. • Assign weights to meta-structures. [ABSTRACT FROM AUTHOR]

Network embedding can effectively dig out potentially useful information and discover the relationships and rules which exist in the data, that has attracted increasing attention in many real-world applications. The goal of network embedding is to map high-dimensional and sparse networks into low-dimensional and dense vector representations. In this paper, we propose a network embedding method based on high-degree penalty and adaptive negative sampling (NEPS). First, we analyze the problem of imbalanced node training in random walk and propose an indicator base on high-degree penalty, which can control the random walk and avoid over-sampling high-degree neighbor node. Then, we propose a two-stage adaptive negative sampling strategy, which can dynamically obtain negative samples suitable for the current training according to the training stage to improve training effect. By comparing with seven well-known network embedding algorithms on eight real-world data sets, experiments show that the NEPS has good performance in node classification, network reconstruction and link prediction. The code is available at: https://github.com/Andrewsama/NEPS-master. [ABSTRACT FROM AUTHOR]

Most existing network embedding based anomalous link detection methods regard network embedding and anomalous link detection as two independent tasks. However, removing anomalous links from the original network can reduce the data noise, thus hopefully improving the performance of network embedding models and anomalous link detection. In this paper, we propose an Anomaly Aware Network Embedding (AANE) framework by simultaneously learning node embedding and detecting anomalous links in a unified way. To instantiate the AANE framework, we propose a heuristic anomalous link selection based model AANE-H and an embedding disentangling based model AANE-D on Graph Auto-Encoder (GAE). In AANE-H, we adopt an anomalous link selector to iteratively select significant anomalous links based on a heuristic rule during model training, while in AANE-D the normal and anomalous links are generated by disentangled normal and anomalous embedding respectively. For the evaluation purpose, we propose a heuristic anomalous link generation algorithm to inject synthetic anomalous links into six real-world network datasets used in our experiments. Experimental results show that AANE outperforms both the state-of-the-art network embedding models and anomalous node detection models in terms of anomalous link detection performance. As a general network embedding model, AANE can also improve other downstream tasks like node classification. [ABSTRACT FROM AUTHOR]

Random-walk-based graph representation methods have been widely applied in User Identity Linkage (UIL) tasks, which links overlapping users between two different social networks. It can help us to obtain more comprehensive portraits of criminals, which is helpful for improving cyberspace governance. Yet, random walk generates a large number of repeating sequences, causing unnecessary computation and storage overhead. This paper proposes a novel method called Edge-Removing Walk (ERW) that can replace random walk in random-walk-based models. It removes edges once they are walked in a walk round to capture the l − h o p features without repetition, and it walks the whole graph for several rounds to capture the different kinds of paths starting from a specific node. Experiments proved that ERW can exponentially improve the efficiency for random-walk-based UIL models, even maintaining better performance. We finally generalize ERW into a general User Identity Linkage framework called ERW-UIL and verify its performance. [ABSTRACT FROM AUTHOR]

Network embedding, as an approach to learning low-dimensional representations of nodes, has been proved extremely useful in many applications, e.g., node classification and link prediction. Unfortunately, existing network embedding models are vulnerable to random or adversarial perturbations, which may degrade the performance of network embedding when being applied to downstream tasks. To achieve robust network embedding, researchers introduce adversarial training to regularize the embedding learning process by training on a mixture of adversarial examples and original examples. However, existing methods generate adversarial examples heuristically, failing to guarantee the imperceptibility of generated adversarial examples, and thus limit the power of adversarial training. In this paper, we propose a novel method Identity-Preserving Adversarial Training (IPAT) for network embedding, which generates imperceptible adversarial examples with explicit identity-preserving regularization. We formalize such identity-preserving regularization as a multi-class classification problem where each node represents a class, and we encourage each adversarial example to be discriminated as the class of its original node. Extensive experimental results on real-world datasets demonstrate that our proposed IPAT method significantly improves the robustness of network embedding models and the generalization of the learned node representations on various downstream tasks. [ABSTRACT FROM AUTHOR]

Network data is ubiquitous, such as telecommunication, transport systems, online social networks, protein-protein interactions, etc. Since the huge scale and the complexity of network data, former machine learning system tried to understand network data arduously. On the other hand, thought of multi-granular cognitive computation simulates the problem-solving process of human brains. It simplifies the complex problems and solves problems from the easier to harder. Therefore, the application of multi-granularity problem-solving ideas or methods to deal with network data mining is increasingly adopted by researchers either intentionally or unintentionally. This paper looks into the domain of network representation learning (NRL). It systematically combs the research work in this field in recent years. In this paper, it is discovered that in dealing with the complexity of the network and pursuing the efficiency of computing resources, the multi-granularity solution becomes an excellent path that is hard to go around. Although there are several papers about survey of NRL, to our best knowledge, we are the first to survey the NRL from the perspective of multi-granular computing. This paper proposes the challenges that NRL meets. Furthermore, the feasibility of solving the challenges of NRL with multi-granular computing methodologies is analyzed and discussed. Some potential key scientific problems are sorted out and prospected in applying multi-granular computing for NRL research. [ABSTRACT FROM AUTHOR]

Complex networks are widely used to represent an abundance of real-world relations ranging from social networks to brain networks. Inferring missing links or predicting future ones based on the currently observed network is known as the link prediction task. Recent network embedding based link prediction algorithms have demonstrated ground-breaking performance on link prediction accuracy. Those algorithms usually apply node attributes as the initial feature input to accelerate the convergence speed during the training process. However, they do not take full advantage of node feature information. In this paper, besides applying feature attributes as the initial input, we make better utilization of node attribute information by building attributable networks and plugging attributable networks into some typical link prediction algorithms and name this algorithm Attributive Graph Enhanced Embedding (AGEE). AGEE is able to automatically learn the weighting trades-off between the structure and the attributive networks. Numerical experiments show that AGEE can improve the link prediction accuracy by around 3% compared with SEAL, Variational Graph AutoEncoder (VGAE), and node2vec.

Abstract Metawebs (networks of potential interactions within a species pool) are a powerful abstraction to understand how large‐scale species interaction networks are structured. Because metawebs are typically expressed at large spatial and taxonomic scales, assembling them is a tedious and costly process; predictive methods can help circumvent the limitations in data deficiencies, by providing a first approximation of metawebs. One way to improve our ability to predict metawebs is to maximize available information by using graph embeddings, as opposed to an exhaustive list of species interactions. Graph embedding is an emerging field in machine learning that holds great potential for ecological problems. Here, we outline how the challenges associated with inferring metawebs line‐up with the advantages of graph embeddings; followed by a discussion as to how the choice of the species pool has consequences on the reconstructed network, specifically as to the role of human‐made (or arbitrarily assigned) boundaries and how these may influence ecological hypotheses.

Abstract Network embedding has been extensively used in several practical applications and achieved great success. However, existing studies mainly focus on single task or single view and cannot obtain deeper relevant information for accomplishing tasks. In this paper, a novel approach is proposed to address the problem of insufficient information consideration in network embedding, which is termed multi-task-oriented adaptive dual-channel graph convolutional network (TAD-GCN). We firstly use kNN graph construction method to generate three views for each network dataset. Then, the proposed TAD-GCN contains dual-channel GCN which can extract the specific and shared embeddings from multiple views simultaneously, and attention mechanism is adopted to fuse them adaptively. In addition, we design similarity constraint and difference constraint to further enhance their semantic similarity and ensure that they capture the different information. Lastly, a multi-task learning module is introduced to solve multiple tasks simultaneously and optimize the model with its losses. The experimental results demonstrate that our model TAD-GCN not only completes multiple downstream tasks at the same time, but also achieves excellent performance compared with eight state-of-the-art methods.

For networking of big data applications, an essential issue is how to represent networks in vector space for further mining and analysis tasks, e.g., node classification, clustering, link prediction, and visualization. Most existing studies on this subject mainly concentrate on monoplex networks considering a single type of relation among nodes. However, numerous real-world networks are naturally composed of multiple layers with different relation types; such a network is called a multiplex network. The majority of existing multiplex network embedding methods either overlook node attributes, resort to node labels for training, or underutilize underlying information shared across multiple layers. In this paper, we propose Multiplex Network Infomax (MNI), an unsupervised embedding framework to represent information of multiple layers into a unified embedding space. To be more specific, we aim to maximize the mutual information between the unified embedding and node embeddings of each layer. On the basis of this framework, we present an unsupervised network embedding method for attributed multiplex networks. Experimental results show that our method achieves competitive performance on not only node-related tasks, such as node classification, clustering, and similarity search, but also a typical edge-related task, i.e., link prediction, at times even outperforming relevant supervised methods, despite that MNI is fully unsupervised.

Learning representation of networks is increasingly important in social network analysis. Network embedding techniques have a wide range of applications in various downstream tasks, such as node classification and clustering. However, many existing approaches focus on representing each node in a network as a single vector, which is often not practical. In this paper, we propose a multiple role-based social network embedding, called mr2vec , by identifying multiple roles of each node through overlapping clustering and encoding each of them to a vector. Our proposed algorithm transforms a given network into the persona graph using overlapping clustering. Each pair of nodes and roles is mapped to a persona node, enabling us to learn multi-role-based embedding using conventional network embedding techniques. Moreover, we develop a new loss function to have close vector representations when the similarity between the nodes in a persona graph is high based not only on the structural information of node, but also on the role information of nodes. We conducted link prediction experiments within the network to demonstrate the effectiveness of representation learning that takes into account the multiple role information discovered through overlapping clustering of each node. Through extensive experiments on social networks, we have shown that mr2vec achieves higher accuracy than the state-of-the-art algorithms for single- or multi-aspect network embedding. • Learning multiple representations by considering the multiple roles of nodes. • Constructing a persona network that integrates both local and global structures. • Developing a new loss function that takes into account the diverse roles of nodes. • Achieving high performance in link prediction using real-world network datasets. [ABSTRACT FROM AUTHOR]

Metawebs (networks of potential interactions within a species pool) are a powerful abstraction to understand how large‐scale species interaction networks are structured.Because metawebs are typically expressed at large spatial and taxonomic scales, assembling them is a tedious and costly process; predictive methods can help circumvent the limitations in data deficiencies, by providing a first approximation of metawebs.One way to improve our ability to predict metawebs is to maximize available information by using graph embeddings, as opposed to an exhaustive list of species interactions. Graph embedding is an emerging field in machine learning that holds great potential for ecological problems.Here, we outline how the challenges associated with inferring metawebs line‐up with the advantages of graph embeddings; followed by a discussion as to how the choice of the species pool has consequences on the reconstructed network, specifically as to the role of human‐made (or arbitrarily assigned) boundaries and how these may influence ecological hypotheses. [ABSTRACT FROM AUTHOR]

Recommendation plays an important role in satisfying users􀆳 needs of information and solving information overload. Heterogeneous information network contains rich semantics and provides a new way for recommendation optimization. Based on the research of recommendation on heterogeneous information network at home and abroad, this paper conducts bibliometric analysis and visual analysis with SATI, Ucinet, NetDraw, and SPSS to obtain the current research focuses and progress. According to the clustering results of literature keywords, the previous research are mainly based on clustering, random walk, meta-path, matrix factorization, network embedding algorithms, and applied in academic research, points of interest, Web services, social friends, patent trading, news and other recommendation scenarios. There is still a large development space for recommendation research based on heterogeneous information network. Future research can be carried out in dynamic recommendation, deep network representation learning and wider applications. [ABSTRACT FROM AUTHOR]

Given a graph G where each node is associated with a set of attributes, attributed network embedding (ANE) maps each node v ∈ G to a compact vector X v , which can be used in downstream machine learning tasks. Ideally, X v should capture node v's affinity to each attribute, which considers not only v's own attribute associations, but also those of its connected nodes along edges in G. It is challenging to obtain high-utility embeddings that enable accurate predictions; scaling effective ANE computation to massive graphs with millions of nodes pushes the difficulty of the problem to a whole new level. Existing solutions largely fail on such graphs, leading to prohibitive costs, low-quality embeddings, or both. This paper proposes PANE , an effective and scalable approach to ANE computation for massive graphs that achieves state-of-the-art result quality on multiple benchmark datasets, measured by the accuracy of three common prediction tasks: attribute inference, link prediction, and node classification. PANE obtains high scalability and effectiveness through three main algorithmic designs. First, it formulates the learning objective based on a novel random walk model for attributed networks. The resulting optimization task is still challenging on large graphs. Second, PANE includes a highly efficient solver for the above optimization problem, whose key module is a carefully designed initialization of the embeddings, which drastically reduces the number of iterations required to converge. Finally, PANE utilizes multi-core CPUs through non-trivial parallelization of the above solver, which achieves scalability while retaining the high quality of the resulting embeddings. The performance of PANE depends upon the number of attributes in the input network. To handle large networks with numerous attributes, we further extend PANE to PANE + + , which employs an effective attribute clustering technique. Extensive experiments, comparing 10 existing approaches on 8 real datasets, demonstrate that PANE and PANE + + consistently outperform all existing methods in terms of result quality, while being orders of magnitude faster. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Technology Economics is the property of Chinese Society of Technology Economics and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)