In this work, we present a novel approach for link prediction on heterogeneous networks – networks that accommodate multiple types of nodes as well as multiple types of relations among them. Specifically, we propose a multi-view network representation learning framework to incorporate structural intuitions from the underlying graph and enrich the relational representations for link prediction. The method relies on the metapath view, the community view, and the subgraph view between a source and target node pair whose linkage is to be predicted. Furthermore, our proposed model leverages a relation-aware attention mechanism to aggregate the candidate contexts in a principled way. Empirically, we demonstrate that the proposed architecture outperforms state-of-the-art transductive and inductive methods in link prediction by a significant margin. A detailed ablation study and attention weight visualizations suggest that the chosen views are complementary and useful to predict links robustly.

Video traffic demands for vehicle infotainment services have rapidly increased in recent years. To reduce bandwidth costs and delays, content distributors (CD) use edge vehicles to distribute video contents over vehicular networks. CDs usually offer multiple subscription tiers of varying video qualities for their clients. A CD’s business goal is to make a minimal edge vehicle selection to maximize Internet traffic offloading while satisfying heterogeneous video quality requirements of clients belonging to different subscription tiers. This work proposes an optimum cost-effective solution for edge vehicle selection for delay-constrained live video streaming over vehicular networks with multi-tier bitrate agreements. The edge vehicle selection problem is formulated as the minimum set covering problem for optimum results, and a distributed approximation solution is proposed to increase scalability. Simulation results show that the proposed solution is cost-effective for CDs and saves significant control message overhead while satisfying delay constraints and multi-tier bitrate agreements.

Abstract Word-level language identification is an essential prerequisite for extracting useful information from code-mixed social media content. Previous studies in word-level language identification show two important observations. First, the local context is an important indicator of the language of a word when a word is valid in multiple languages. Second, considering the word in isolation from its context leads to more effective language classification when a word is borrowed or embedded into sentences of other languages. In this paper, we propose a framework for language identification that makes use of a dynamic switching mechanism for effective language classification of both words that are borrowed or embedded from other languages as well as words that are valid in multiple languages. For a given input, the proposed switching mechanism makes a dynamic decision to bias its prediction either towards the prediction obtained by the contextual information or that obtained by the word in isolation. In contrast to existing studies that rely upon large amounts of annotated data for robust performance in a multilingual environment, the proposed approach uses minimal annotated resources and no external resources, making it easily extendible to newer languages. Evaluation over a corpus of transliterated Facebook comments shows that the proposed approach outperforms its baseline counterparts: classification based on the contextual information, classification based on the word in isolation, as well as an ensemble of the two classifiers.

The volume of video traffic as infotainment service over vehicular ad hoc network (VANET) has rapidly increased for past few years. Providing video streaming as VANET infotainment service is very challenging because of high mobility and heterogeneity of vehicular networks. While the number of mobile gateways (vehicles connected to the Internet) needs to be minimized to reduce service cost, the streaming quality also needs to be satisfactory for end-users. Existing works either focus on gateway minimization or focus on enhancing user satisfaction. We propose a video streaming solution, namely, DGSVS, which does gateway minimization with the constrained time data delivery to end-users. We formulate our constrained gateway minimization problem as minimum set covering (MSC) problem and solve with a distributed approximation method for MSC. We assume that only a subset of vehicles in VANET run DGSVS application. Therefore, instead of application layer cooperation for gateway-client association, network layer cooperation is proposed. We propose a novel multicast protocol DSS-CAST for this purpose, which is specialized in streaming data distribution for dynamic scenarios. We compare the performance of DGSVS with other existing protocols and found that DGSVS is most effective in service cost minimization while it is able to achieve competitive QoE performance.

Multiplex networks are complex graph structures in which a set of entities are connected to each other via multiple types of relations, each relation representing a distinct layer. Such graphs are used to investigate many complex biological, social, and technological systems. In this work, we present a novel semi-supervised approach for structure-aware representation learning on multiplex networks. Our approach relies on maximizing the mutual information between local node-wise patch representations and label correlated structure-aware global graph representations to model the nodes and cluster structures jointly. Specifically, it leverages a novel cluster-aware, node-contextualized global graph summary generation strategy for effective joint-modeling of node and cluster representations across the layers of a multiplex network. Empirically, we demonstrate that the proposed architecture outperforms state-of-the-art methods in a range of tasks: classification, clustering, visualization, and similarity search on seven real-world multiplex networks for various experiment settings.

Data centers are over provisioned to handle traffic surges. However, various studies of data center traffic regarded this upsurge as a rare event and shown that the resources of data center are underutilized. Hence, there is a need to conserve energy by providing control knobs that turn the resources of data center on/off according to the traffic condition. In this paper, we study the ability of 4-4, 1-4 data center network design for energy conservation. Given a set of current flows, we propose a greedy approach to find sub-network that can accommodate current set of flows and switch off rest of the network to conserve energy. We simulated the proposed method using network-simulator (NS3) and compared the performance with Elastic Tree. The simulation results shown performance of 4-4, 1-4 design is better than Elastic Tree in terms of energy conservation, packet delivery ratio and delay. As an extension, we also compared the support from routing protocols: source routing and location based routing, in designing energy conservation methods.

The increasing complexity and sophistication of the applications deployed on Data Center Network (DCN) demanded new features and greater performance from the DCN. This resulted in many designs addressing various challenges such as cost, performance, reliability, scalability, security and energy. One major challenge a designer often faces is the realization of their proposed design or realization of the existing designs for comparison. Although proto-typing is a better choice but it does have certain limitation and is very complex and expensive. Hence, Simulation is considered as an alternative to the prototyping. In this paper, we present a case study of using Network Simulator 3 (NS3) for realization of various architectures for DCN and study their performance. The information we provide includes realization of the most popular designs for DCN and tools available with NS3 to study their performance. Our effort is to make it easy for a beginner to build popular designs for DCN and study their performance using NS3.

Architecture plays an important role in the design of Data Center Networks (DCN). Designs for DCN are to be scalable, robust and efficient. There are many designs proposed in recent years where routing is performed using location information of the server. These designs solve scalability problem of IP and Ethernet with additional overhead of determining and maintaining location information of the server. In this paper, we propose a new architecture named, four-four, one-four architecture (4-4, 1-4 Architecture) that has evolved from the format of IP address. Addressing and interconnection used in our design eliminate the complexities in the previous designs such as constructing addresses based on location information. Two major benefits - reduction of entries in routing table and routing based on location information - are achieved through our design with minimum complexity. The first advantage is due to the design combining servers with common prefixes at each level using hierarchical format of IP address. The second advantage is due to the IP address assigned to servers and switches acting as location indicator. This helps in faster routing where packet forwarding is done based on the bits of IP addresses. Further, the proposed architecture meets major design requirements for data center networks such as load balancing and energy conservation.

Tuple Space based Mobile Middleware (TSMM) is developed to tackle emerging dynamics in underlying infrastructure. Its coordination medium, viz. tuple space model, has multiple inherent decoupled behaviors to coordinate interactions between different agents of supported applications. This paper focuses on one such decoupling quality, synchronization decoupling, by which agents are decoupled from their invoked primitives. This decoupling ability is accomplished by including reactivity in tuple space model. Reactive tuple space model also enables development of robust and flexible applications for TSMM. However, in existing TSMM, this decoupling behavior has not been fully attained, as reactions are themselves coupled with invoked primitives, leading to additional network and processing overheads. This paper proposes an improved reactive tuple space model for TSMM with enhanced decoupling behavior of reactions to achieve robustness for supported applications. Proposed tuple space model is also formally modeled and verified by notation and proof logic of UNITY model.

Abstract Slicing of concurrent programs is a compute‐intensive task. To speed up the slicing process, we have developed a parallel algorithm. For this purpose we used the concurrent control flow graph (CCFG) as the intermediate representation. We used a network of communicating processes to develop our parallel algorithm. We have implemented our parallel algorithm and the experimental results appear promising. Copyright © 2004 John Wiley & Sons, Ltd.

We present a method for computing static slices of concurrent programs in a Unix process environment. As a part of our methodology, we introduce the notion of a Concurrent Program Dependence Graph (CPDG). A CPDG represents various aspects of concurrent programs in a hierarchical fashion. This hierarchical representation lets us compute static slices of programs at different levels of abstraction. Based on our methodology, we have implemented a static slicing tool supporting an option to view slices of programs at different levels of details. Experience with our implementation shows that this approach helps the user get a better understanding of the behavior of concurrent programs. Copyright © 2000 John Wiley & Sons, Ltd.