Network Embedding Algorithm Research Articles

Attributed network embedding via subspace discovery

Network embedding aims to learn a latent, low-dimensional vector representations of network nodes, effective in supporting various network analytic tasks. While prior arts on network embedding focus primarily on preserving network topology structure to learn node representations, recently proposed attributed network embedding algorithms attempt to integrate rich node content information with network topological structure for enhancing the quality of network embedding. In reality, networks often have sparse content, incomplete node attributes, as well as the discrepancy between node attribute feature space and network structure space, which severely deteriorates the performance of existing methods. In this paper, we propose a unified framework for attributed network embedding-attri2vec-that learns node embeddings by discovering a latent node attribute subspace via a network structure guided transformation performed on the original attribute space. The resultant latent subspace can respect network structure in a more consistent way towards learning high-quality node representations. We formulate an optimization problem which is solved by an efficient stochastic gradient descent algorithm, with linear time complexity to the number of nodes. We investigate a series of linear and non-linear transformations performed on node attributes and empirically validate their effectiveness on various types of networks. Another advantage of attri2vec is its ability to solve out-of-sample problems, where embeddings of new coming nodes can be inferred from their node attributes through the learned mapping function. Experiments on various types of networks confirm that attri2vec is superior to state-of-the-art baselines for node classification, node clustering, as well as out-of-sample link prediction tasks. The source code of this paper is available at https://github.com/daokunzhang/attri2vec.

SepNE: Bringing Separability to Network Embedding

Many successful methods have been proposed for learning low dimensional representations on large-scale networks, while almost all existing methods are designed in inseparable processes, learning embeddings for entire networks even when only a small proportion of nodes are of interest. This leads to great inconvenience, especially on super-large or dynamic networks, where these methods become almost impossible to implement. In this paper, we formalize the problem of separated matrix factorization, based on which we elaborate a novel objective function that preserves both local and global information. We further propose SepNE, a simple and flexible network embedding algorithm which independently learns representations for different subsets of nodes in separated processes. By implementing separability, our algorithm reduces the redundant efforts to embed irrelevant nodes, yielding scalability to super-large networks, automatic implementation in distributed learning and further adaptations. We demonstrate the effectiveness of this approach on several real-world networks with different scales and subjects. With comparable accuracy, our approach significantly outperforms state-of-the-art baselines in running times on large networks.

Robust Negative Sampling for Network Embedding

Many recent network embedding algorithms use negative sampling (NS) to approximate a variant of the computationally expensive Skip-Gram neural network architecture (SGA) objective. In this paper, we provide theoretical arguments that reveal how NS can fail to properly estimate the SGA objective, and why it is not a suitable candidate for the network embedding problem as a distinct objective. We show NS can learn undesirable embeddings, as the result of the “Popular Neighbor Problem.” We use the theory to develop a new method “R-NS” that alleviates the problems of NS by using a more intelligent negative sampling scheme and careful penalization of the embeddings. R-NS is scalable to large-scale networks, and we empirically demonstrate the superiority of R-NS over NS for multi-label classification on a variety of real-world networks including social networks and language networks.

Personalized Question Routing via Heterogeneous Network Embedding

Question Routing (QR) on Community-based Question Answering (CQA) websites aims at recommending answerers that have high probabilities of providing the “accepted answers” to new questions. The existing question routing algorithms simply predict the ranking of users based on query content. As a consequence, the question raiser information is ignored. On the other hand, they lack learnable scoring functions to explicitly compute ranking scores.To tackle these challenges, we propose NeRank that (1) jointly learns representations of question content, question raiser, and question answerers by a heterogeneous information network embedding algorithm and a long short-term memory (LSTM) model. The embeddings of the three types of entities are unified in the same latent space, and (2) conducts question routing for personalized queries, i.e., queries with two entities (question content, question raiser), by a convolutional scoring function taking the learned embeddings of all three types of entities as input. Using the scores, NeRank routes new questions to high-ranking answerers that are skillfulness in the question domain and have similar backgrounds to the question raiser.Experimental results show that NeRank significantly outperforms competitive baseline question routing models that ignore the raiser information in three ranking metrics. In addition, NeRank is convergeable in several thousand iterations and insensitive to parameter changes, which prove its effectiveness, scalability, and robustness.

Outlier Aware Network Embedding for Attributed Networks

Attributed network embedding has received much interest from the research community as most of the networks come with some content in each node, which is also known as node attributes. Existing attributed network approaches work well when the network is consistent in structure and attributes, and nodes behave as expected. But real world networks often have anomalous nodes. Typically these outliers, being relatively unexplainable, affect the embeddings of other nodes in the network. Thus all the downstream network mining tasks fail miserably in the presence of such outliers. Hence an integrated approach to detect anomalies and reduce their overall effect on the network embedding is required.Towards this end, we propose an unsupervised outlier aware network embedding algorithm (ONE) for attributed networks, which minimizes the effect of the outlier nodes, and hence generates robust network embeddings. We align and jointly optimize the loss functions coming from structure and attributes of the network. To the best of our knowledge, this is the first generic network embedding approach which incorporates the effect of outliers for an attributed network without any supervision. We experimented on publicly available real networks and manually planted different types of outliers to check the performance of the proposed algorithm. Results demonstrate the superiority of our approach to detect the network outliers compared to the state-of-the-art approaches. We also consider different downstream machine learning applications on networks to show the efficiency of ONE as a generic network embedding technique. The source code is made available at https://github.com/sambaranban/ONE.

Image classification algorithm based on social network metadata

The images in online social networks usually carry a large amount of social network metadata, such as labels, users, picture groups, locations and comments. These social network metadata, which include rich image semantic information, can help users distinguish the content of imagesand communicate with each other. An image classification algorithm based on multiple social networks, MSNet, is proposed in this paper. Firstly, MSNet collects the social network metadata of images and constructs relationship networks of images based on the metadata. Then, a network embedding algorithm is used to learn the representation vectors of images in each network. Finally, a neural network classifier is trained to classify the images by using the visual features and network representation of images. The experimental results on PASCAL、MIR、CLEF and NUS image data sets show the superiority performance of MSNet in comparison with CNN-neighbor and kernel canonical correlation analysis (KCCA).

Predicting drug–disease associations by network embedding and biomedical data integration

PurposeThe traditional drug development process is costly, time consuming and risky. Using computational methods to discover drug repositioning opportunities is a promising and efficient strategy in the era of big data. The explosive growth of large-scale genomic, phenotypic data and all kinds of “omics” data brings opportunities for developing new computational drug repositioning methods based on big data. The paper aims to discuss this issue.Design/methodology/approachHere, a new computational strategy is proposed for inferring drug–disease associations from rich biomedical resources toward drug repositioning. First, the network embedding (NE) algorithm is adopted to learn the latent feature representation of drugs from multiple biomedical resources. Furthermore, on the basis of the latent vectors of drugs from the NE module, a binary support vector machine classifier is trained to divide unknown drug–disease pairs into positive and negative instances. Finally, this model is validated on a well-established drug–disease association data set with tenfold cross-validation.FindingsThis model obtains the performance of an area under the receiver operating characteristic curve of 90.3 percent, which is comparable to those of similar systems. The authors also analyze the performance of the model and validate its effect on predicting the new indications of old drugs.Originality/valueThis study shows that the authors’ method is predictive, identifying novel drug–disease interactions for drug discovery. The new feature learning methods also positively contribute to the heterogeneous data integration.

VNE-TD: A virtual network embedding algorithm based on temporal-difference learning

Recently, network virtualization is considered as a promising solution for the future Internet which can help to overcome the resistance of the current Internet to fundamental changes. The problem of embedding Virtual Networks (VN) in a Substrate Network (SN) is the main resource allocation challenge in network virtualization. The major challenge of the Virtual Network Embedding (VNE) problem lies in the contradiction between making online embedding decisions and pursuing a long-term objective. Most previous works resort to balancing the SN workload with various methods to deal with this contradiction. Rather than passive balancing, we try to overcome it by learning actively and making online decisions based on previous experiences. In this article, we model the VNE problem as Markov Decision Process (MDP) and develop a neural network to approximate the value function of VNE states. Further, a VNE algorithm based on Temporal-Difference Learning (one kind of Reinforcement Learning methods), named VNE-TD, is proposed. In VNE-TD, multiple embedding candidates of node-mapping are generated probabilistically, and TD Learning is involved to evaluate the long-run potential of each candidate. Extensive simulation results show that VNE-TD outperforms previous algorithms significantly in terms of both block ratio and revenue.

DYVINE: Fitness-Based Dynamic Virtual Network Embedding in Cloud Computing

Virtual network embedding (VNE) is the process of embedding the set of interconnected virtual machines onto the set of interconnected physical servers (PSs) in the cloud computing environment. The level of complexity of VNE problem increases when a large number of virtual machines with a set of resource demand need to be embedded onto a network of thousands of PSs. The key challenge of VNE is the efficient mapping of virtual networks (VNs), which may have dynamic resource demands. Existing solutions mainly emphasize on the embedding of static VN resulting in poor resource utilization and very low acceptance rate. To tackle such level of complexity in VNE, a fitness-based dynamic virtual network embedding (DYVINE) algorithm is proposed with the goal to maximize the resource utilization by maximizing the acceptance rate. Local and global fitness values of the virtual machines and VN, respectively, are used to utilize the maximum amount of physical resources. The proposed VNE algorithm allows the VN to be dynamic, which indicates that the structure and resource demand can be changed during its execution time. Furthermore, in order to reduce the embedding time in each time slot, a set of PSs is selected to host the VN instead of considering thousands of PSs, which may significantly increase the embedding time. The proposed embedding mechanism is evaluated through extensive simulation and is compared with similar existing embedding algorithms, which outperforms over others.

Security Aware Virtual Network Embedding Algorithm Using Information Entropy TOPSIS

Network virtualization is an effective manner to address the ossification issue of Internet architecture. Virtual network embedding is one of the most critical techniques in network virtualization environments. Several security problems about virtual network embedding are introduced due to the fact that virtual network embedding adds a virtual layer into the internet architecture. In this paper, we proposed an approach for security aware virtual network embedding called SA-VNE to address the security problems in virtual network embedding process. Firstly, we use the information entropy TOPSIS method to rank the importance of substrate nodes with an aim to choose the most appropriate substrate node for accommodating the virtual node. Secondly, we use the shortest path algorithm to perform the link mapping process. Simulation results demonstrated that our proposed SA-VNE algorithm behaves better that those of state-of-the-art existing security aware virtual network embedding algorithms in terms of the long-term average revenue, the long-term average VN acceptance ratio, the long-term average revenue to cost ratio and the running time.

A PSO based multi-domain virtual network embedding approach

This paper proposed a multi-domain virtual network embedding algorithm based on multi-controller SDN architecture. The local controller first selects candidate substrate nodes for each virtual node in the domain. Then the global controller abstracts substrate network topology based on the candidate nodes and boundary nodes of each domain, and applies Particle Swarm Optimization Algorithm on it to divide virtual network requests. Each local controller then embeds the virtual nodes of the divided single-domain virtual network requests in the domain, and cooperates with other local controllers to embed the inter-domain virtual links. Simulation experimental results show that the proposed algorithm has good performance in reducing embedding cost with good stability and scalability.

A Self-Adaptive Virtual Network Embedding Algorithm Based on Software-Defined Networks

Network virtualization provides a promising tool to allow multiple virtual networks (VNs) to run on a shared substrate network (SN) simultaneously. VN embedding (VNE) is one of the key technologies of network virtualization. The main goal of VNE is to effectively map VN requests to the SN, which is efficiently utilizes the network resources. The emergence of software defined networks provides a platform for network virtualization to be used and promoted. In a real environment, the resource requirements of tenants are generally different. A single VN mapping algorithm can not effectively handle the multi-demand problem of tenants. We propose a self-adaptive VNE algorithm. VN requests are divided into different types by an adaptive algorithm, we use an integer linear programming formulation to solve VNE problem. This paper considers three different types of VN requests. Type 1 VN requests for high bandwidth requirements, type 2 VN requests for low latency requirements, and type 3 VN requests for high bandwidth requirements and latency requirements. The simulation results show that the virtual network embedding algorithm proposed in this paper can make full use of the SN resources and improve the overall revenue, while effectively dealing with the multi-demand problem of tenants.

ACE: Ant Colony Based Multi-Level Network Embedding for Hierarchical Graph Representation Learning

As a popularly used technique for feature learning in graphs, network embedding aims to represent each node as a low-dimensional vector to support efficient graph analytic tasks, such as node classification, link prediction, and visualization. The key to this representation method is that the embedding vector of a node should preserve its properties in the graph as much as possible. Most traditional network embedding algorithms only consider the local neighborhood as the context to build the node representation and fail to capture the important hierarchical clustering property ubiquitous in real-world graphs. To solve this problem, we propose ACE, a novel network embedding method, to preserve the features of hierarchical clustering structures. ACE works by using an ant colony-based graph coarsening algorithm to group the nodes according to their relationship to achieve a multi-level clustering pyramid of the input graph. Then, we generate the embedding vectors from multiple layers of the graph pyramid and blend these multi-level vectors into the final representation of nodes based on the PCA dimension reduction algorithm. We demonstrate the effectiveness of ACE over state-of-the-art network embedding algorithms on the node classification tasks in several real-world graph datasets. The experiments show that ACE is easy to be integrated with other network embedding algorithms, such as DeepWalk, Line, node2vec, and SDNE, to significantly improve their performance by up to 22% on Macro F1. The source code and the datasets used in this paper are available on Github ( https://github.com/so-link/ACE_Embedding ).

DeepAttr: Inferring Demographic Attributes via Social Network Embedding

Online Social Networks (OSNs) (e.g., Facebook, Twitter, and Tecent QQ) are popular platforms for people to share information online. The providers of OSNs typically leverage demographic attributes of users in OSNs to perform business behaviors like personalized commodity recommendation, advertisement delivery, etc. Thus, understanding demographic attributes of users in OSNs is of great importance. However, some users refuse to reveal their demographic information due to privacy concerns. Therefore, inferring demographic attributes of users in OSNs with public information is an attractive topic for researchers. Most existing methods mainly focus on individual attribute inference without taking the relationship among attributes into account, which results in low precisions. In this paper, we propose a novel approach, called DeepAttr, to infer users' multiple attributes simultaneously. DeepAttr leverages existing network embedding algorithm to learn the social embedding for each user. Meanwhile, it encodes multiple attributes into structured attribute vectors which are treated as class labels. The core component of DeepAttr is a multi-layer fully connected deep neural network which captures complex nonlinear mapping between the users' social embeddings and the structured attribute vectors. Extensive experiments on a real-world dataset demonstrate that DeepAttr outperforms existing models in both single-attribute inference and multiple-attribute inference as to macro-Precision, macro-Recall, and macro-F1.

Virtual Network Embedding Algorithm via Diffusion Wavelet

The great success of the Internet has promoted the development of digital industries and increased the demand for communication bandwidth. For example, ultrahigh-definition videos and vehicle networks require fast bandwidth speed and increase network connection density, respectively. High-bandwidth and high-density parallel communication drive the rapid development of network virtualization and 5G/6G technology. In a network virtualization environment, this new demand also brings new link resource allocation difficulties in existing substrate networks. To solve this far-reaching problem, this paper proposes a virtual network embedding algorithm via diffusion wavelet (VNE_DW), which is an unsupervised structure learning algorithm. Through the diffusion wavelet, the topology structure of nodes, connection density, and link volume among the nodes are comprehensively evaluated. Nodes that facilitate the link mapping success rate are preferentially selected. Experimental results demonstrate that the mapping success rate and revenue-cost ratio of VNE_DW outperform other state-of-the-art algorithms with high bandwidth and density.

Enhancing Attributed Network Embedding via Similarity Measure

Network embedding aims to represent network structural and attributed information with low-dimensional vectors, which has been demonstrated to be beneficial for many network analysis tasks, such as link prediction, node classification and visualization. However, nodes in networks are commonly associated with rich contents, which are facilitated to characterize the properties of nodes. Most existing attributed network embedding algorithms tend to learn attribute representations separated from structure representations, which require a subsequent processing of combination. Besides, these traditional approaches ignore the potential high-order proximity introduced by attributes. Motivated by this, we investigate how structures and attributes can be captured simultaneously and introduce similarity measure to preserve high-order proximity in an attributed network. In this paper, we propose a novel attributed network embedding framework, Similarity Enhancing Attributed Network Embedding (SEANE), which jointly preserves structural and attributed information, and adopts similarity measure to enhance the node embedding. We evaluate our proposed framework by using four real-world datasets on link prediction, node classification and nearest nodes searching. The experimental results demonstrate the outperformance of SEANE on link prediction and node classification tasks.

Virtual Network Embedding in Fiber-Wireless Access Networks for Resource-Efficient IoT Service Provisioning

The rise of Industry 4.0 is pushing more machine devices into the connection of the communication network. Internet of Things (IoT) emerges as an enabling technique to support automatic information transmission and the universal interconnection among massive machine devices. Unlike the human type communication (HTC) in the conventional Internet, IoT features machine type communication (HTC) with small data granularity and massive connections, which impose new challenges on the network resource allocation earlier designed for HTC communication. One of the challenging issues is how we can manage the network resource for the IoT service provisioning. As a promising network architecture for the “last-mile” access in 5G communication, Fiber-Wireless (FiWi) access network compromises the high bandwidth capacity of the optical access network and the flexibility of wireless access network. Especially, network virtualization as an emerging technique provides FiWi with the feasibility for the coordinated bandwidth allocation of wireless access network and optical access network. The evolution of FiWi access network should be not only oriented to the technological breakthrough in regular traffic, but also be supportive of the emerging IoT traffic. In this paper, we focus on the resource allocation problem in FiWi access network supporting IoT service. We first analyze the dynamics of regular traffic and propose the traffic prediction method based on Q-Learning. Then, we propose a virtual network embedding algorithm to map exactly the virtual networks of IoT traffic to the idle resource of the virtual networks of regular traffic without degrading the performance of regular traffic. Thus, the proposed algorithm can improve the resource utilization more effectively. The simulation results demonstrate that the proposed algorithm can achieve a higher acceptance ratio and lower migration ratio of IoT traffic.

Node‐Fusion: Topology‐aware virtual network embedding algorithm for repeatable virtual network mapping over substrate nodes

SummaryCloud computing has become a new Internet application model, where network virtualization is recognized as an important technology for allowing multiple heterogeneous virtual networks (VNs) to coexist on a shared substrate network (SN). As demands in cloud computing increase, the scale of VN greatly increases as well, and providing an end‐to‐end SN to embed VNs in terms of scale is difficult. To utilize SN resources fully, we devise a topology‐aware Node‐Fusion algorithm, which is different from the traditional virtual network embedding (VNE) algorithms, for repeatable VNE over substrate nodes problem. We rank the resource of nodes through a novel solution by considering the CPU and bandwidth of adjacent link capacity and the number of adjacent links of each node as resources, and rank a node on the basis of resources. Furthermore, we embed several virtual nodes into the same substrate node together in accordance with Node‐Fusion interconnection value during the node mapping process, which can greatly improve the success ratio of the subsequent link mapping phase. Evaluation results confirm that Node‐Fusion outperforms traditional classical heuristics (Link‐opt, Node‐opt, and ORSTA), which are modified to fit into our model, with regard to acceptance ratio, long‐term revenue, long‐term cost, and revenue‐cost ratio.

Virtual Network Embedding Based on Topology Potential.

To improve the low acceptance ratio and revenue to cost ratio caused by the poor match between the virtual nodes and the physical nodes in the existing virtual network embedding (VNE) algorithms, we established a multi-objective optimization integer linear programming model for the VNE problem, and proposed a novel two-stage virtual network embedding algorithm based on topology potential (VNE-TP). In the node embedding stage, the field theory once used for data clustering was introduced and a node embedding function designed to find the optimal physical node. In the link embedding stage, both the available bandwidth and hops of the candidate paths were considered, and a path embedding function designed to find the optimal path. Extensive simulation results show that the proposed algorithm outperforms other existing algorithms in terms of acceptance ratio and revenue to cost ratio.

An algorithm for fragment-aware virtual network reconfiguration.

In view of the fact that the current online virtual network embedding algorithms do not consider the fragment resources generated in the embedding process deeply enough, resulting in the problem that the acceptance ratio and the revenue to cost ratio are both low, a mathematical model for virtual network reconfiguration is constructed and a heuristic algorithm for fragment-aware virtual network reconfiguration (FA-VNR) is proposed. The FA-VNR algorithm selects the set of virtual nodes to be migrated according to the fragment degrees of the physical nodes, and selects the best virtual node migration scheme according to the reduction of the fragment degrees of the physical nodes as well as the reduction of the embedding cost of the embedded virtual networks. Extensive simulation results show that the proposed FA-VNR algorithm not only can obviously improve the acceptance ratio and the revenue to cost ratio of the current online virtual network embedding algorithm, but also has better optimization effect than the existing virtual network reconfiguration algorithm.