Sparse Network Topologies Research Articles

A Novel Topology Adaptation Strategy for Dynamic Sparse Training in Deep Reinforcement Learning.

Deep reinforcement learning (DRL) has been widely adopted in various applications, yet it faces practical limitations due to high storage and computational demands. Dynamic sparse training (DST) has recently emerged as a prominent approach to reduce these demands during training and inference phases, but existing DST methods achieve high sparsity levels by sacrificing policy performance as they rely on the absolute magnitude of connections for pruning and randomly generating connections. Addressing this, our study presents a generic method that can be seamlessly integrated into existing DST methods in DRL to enhance their policy performance while preserving their sparsity levels. Specifically, we develop a novel method for calculating the importance of connections within the model. Subsequently, we dynamically adjust the sparse network topology by dropping existing connections and introducing new connections based on their respective importance values. Through validation on eight widely used simulation tasks, our method improves two state-of-the-art (SOTA) DST approaches by up to 70% in episode return and average return across all episodes under various sparsity levels.

Structural and functional specializations of human fast-spiking neurons support fast cortical signaling.

Fast-spiking interneurons (FSINs) provide fast inhibition that synchronizes neuronal activity and is critical for cognitive function. Fast synchronization frequencies are evolutionary conserved in the expanded human neocortex despite larger neuron-to-neuron distances that challenge fast input-output transfer functions of FSINs. Here, we test in human neurons from neurosurgery tissue, which mechanistic specializations of human FSINs explain their fast-signaling properties in human cortex. With morphological reconstructions, multipatch recordings, and biophysical modeling, we find that despite threefold longer dendritic path, human FSINs maintain fast inhibition between connected pyramidal neurons through several mechanisms: stronger synapse strength of excitatory inputs, larger dendrite diameter with reduced complexity, faster AP initiation, and faster and larger inhibitory output, while Na+ current activation/inactivation properties are similar. These adaptations underlie short input-output delays in fast inhibition of human pyramidal neurons through FSINs, explaining how cortical synchronization frequencies are conserved despite expanded and sparse network topology of human cortex.

DDQN with Prioritized Experience Replay-Based Optimized Geographical Routing Protocol of Considering Link Stability and Energy Prediction for UANET.

Unmanned aerial vehicles (UAVs) are important equipment for efficiently executing search and rescue missions in disaster or air-crash scenarios. Each node can communicate with the others by a routing protocol in UAV ad hoc networks (UANETs). However, UAV routing protocols are faced with the challenges of high mobility and limited node energy, which hugely lead to unstable link and sparse network topology due to premature node death. Eventually, this severely affects network performance. In order to solve these problems, we proposed the deep-reinforcement-learning-based geographical routing protocol of considering link stability and energy prediction (DSEGR) for UANETs. First of all, we came up with the link stability evaluation indicator and utilized the autoregressive integrated moving average (ARIMA) model to predict the residual energy of neighbor nodes. Then, the packet forward process was modeled as a Markov Decision Process, and according to a deep double Q network with prioritized experience replay to learn the routing-decision process. Meanwhile, a reward function was designed to obtain a better convergence rate, and the analytic hierarchy process (AHP) was used to analyze the weights of the considered factors in the reward function. Finally, to verify the effectiveness of DSEGR, we conducted simulation experiments to analyze network performance. The simulation results demonstrate that our proposed routing protocol remarkably outperforms others in packet delivery ratio and has a faster convergence rate.

Grand Canonical Ensembles of Sparse Networks and Bayesian Inference.

Maximum entropy network ensembles have been very successful in modelling sparse network topologies and in solving challenging inference problems. However the sparse maximum entropy network models proposed so far have fixed number of nodes and are typically not exchangeable. Here we consider hierarchical models for exchangeable networks in the sparse limit, i.e., with the total number of links scaling linearly with the total number of nodes. The approach is grand canonical, i.e., the number of nodes of the network is not fixed a priori: it is finite but can be arbitrarily large. In this way the grand canonical network ensembles circumvent the difficulties in treating infinite sparse exchangeable networks which according to the Aldous-Hoover theorem must vanish. The approach can treat networks with given degree distribution or networks with given distribution of latent variables. When only a subgraph induced by a subset of nodes is known, this model allows a Bayesian estimation of the network size and the degree sequence (or the sequence of latent variables) of the entire network which can be used for network reconstruction.

Joint sparsity-biased variational graph autoencoders

To bring the full benefits of machine learning to defense modeling and simulation, it is essential to first learn useful representations for sparse graphs consisting of both key entities (vertices) and their relationships (edges). Here, we present a new model, the Joint Sparsity-Biased Variational Graph AutoEncoder (JSBVGAE), capable of learning embedded representations of nodes from which both sparse network topologies and node features can be jointly and accurately reconstructed. We show that our model outperforms the previous state of the art on standard link-prediction and node-classification tasks, and achieves significantly higher whole-network reconstruction accuracy, while reducing the number of trained parameters.

Delay-Constrained Topology-Transparent Distributed Scheduling for MANETs

Transparent topology is common in many mobile ad hoc networks (MANETs) such as vehicle ad hoc networks (VANETs), unmanned aerial vehicle (UAV) ad hoc networks, and wireless sensor networks due to their decentralization and mobility nature. There are many existing works on distributed scheduling scheme design for topology-transparent MANETs. Most of them focus on delay-unconstrained settings. However, with the proliferation of real-time applications over wireless communications, it becomes more and more important to support delay-constrained traffic in MANETs. In such applications, each packet has a given hard deadline: if it is not delivered before its deadline, its validity will expire and it will be removed from the system. This feature is fundamentally different from the traditional delay-unconstrained one. In this article, we for the first time investigate distributed scheduling schemes for a topology-transparent MANET to support delay-constrained traffic. We analyze and compare probabilistic ALOHA scheme and deterministic sequence schemes, including the conventional time division multiple access (TDMA), the Galois field (GF) sequence scheme proposed by Chlamtac and Farago and the combination sequence scheme that we propose for a special type of sparse network topology. We use both theoretical analysis and empirical simulations to compare all these schemes and summarize the conditions under which different individual schemes perform best.

Distributed Training of Graph Convolutional Networks

The aim of this work is to develop a fully-distributed algorithmic framework for training graph convolutional networks (GCNs). The proposed method is able to exploit the meaningful relational structure of the input data, which are collected by a set of agents that communicate over a sparse network topology. After formulating the centralized GCN training problem, we first show how to make inference in a distributed scenario where the underlying data graph is split among different agents. Then, we propose a distributed gradient descent procedure to solve the GCN training problem. The resulting model distributes computation along three lines: during inference, during back-propagation, and during optimization. Convergence to stationary solutions of the GCN training problem is also established under mild conditions. Finally, we propose an optimization criterion to design the communication topology between agents in order to match with the graph describing data relationships. A wide set of numerical results validate our proposal. To the best of our knowledge, this is the first work combining graph convolutional neural networks with distributed optimization.

Performance Analysis of IoT Protocol Stack over Dense and Sparse Mote Network Using COOJA Simulator

Internet of Things (IoT) interrelates computing devices, and digital devices with animals or people and they have ability to transfer amounts of data over a network needless of any human to computer interaction. IoT is composed of sensor nodes or motes, servers, and the internet allowing these devices to collect and exchange information. Lossy networks, stressed out connections and high power consumption are the major concerns in IoT. Protocols are used for messages management and among computing nodes and they act as controls for telecommunication medium using radio frequencies. This paper, evaluates the IoT protocols’ stack over on dense network and sparse network topology which include application layer’s protocols such as, MQTT, CoAP and transport layer protocols’ UDP and TCP. Analysis is done on transport layer through RPL protocol and on physical layer using Z1 motes and Sky motes. We utilize Cooja simulator and Contiki operating system for simulations. The performance metrics for the evaluation of protocol stack are power intake, radio responsibility cycle, and average inter-packet arrival time.

A cooperative mobile throwbox-based routing protocol for social-aware delay tolerant networks

In delay tolerant networks (DTNs), nodes have intermittent connectivity patterns due to various factors such as mobility, sparse network topology, and unpredictable movement patterns. In such networks, nodes store messages, carry them along as they move and forward them opportunistically whenever an encounter occurs. In many DTN applications, such as in disaster situations and remote regions with no communications infrastructure, we can envisage a network topology formed by the mobility and contact pattern of people carrying their mobile devices. The social behavior of people (nodes) in Spatio-temporal dimensions governs the formation of such a topology. To improve data dissemination in such networks, several works have proposed to utilize social-aware metrics for the selection of relays. However, existing routing protocols for DTNs may perform poorly in the scenarios in which nodes are localized in multiple small regions, and the topology is sparse such as in co-located villages in remote regions with no communication infrastructure. This work proposes a Cooperative mobile throwbox-based routing protocol (CMTR) for social-aware DTNs. In CMTR, static and mobile ThrowBoxes (TBs) are deployed to improve the efficiency of a DTN. The static and mobile TBs cooperatively relay data to enhance network efficiency. Performance evaluation via simulations in ONE simulator shows that CMTR improves the data delivery ratio, delays, and energy efficiency as compared to Epidemic routing and Bubble Rap routing schemes.

Optimal sparse network topology under sparse control in Laplacian networks

Understanding the effect of network topology on its controllability or control performance helps to explain the emergence and evolution of different topological features in biological or social networks, and to provide principles of designing networked engineering systems. In natural and artificial systems, networks have sparse topology and are controlled by sparse feedback controllers. Therefore, the optimal network topology and controllers are specified by a trade-off between the control performance and the sparsity of the network and the controllers. In this work, by considering Laplacian networks and adopting an H2-optimal control framework that incorporates the network Laplacian matrix as an optimization variable and promotes the network sparsity and controller sparsity, we simultaneously optimize the network topology and the feedback gain matrix through the alternating direction method of multipliers (ADMM). We observe that with increasing sparsity, modular and hierarchical topological features emerge along with distributed and hierarchical control architectures, which is consistent with observations in biological networks and allows an evolutionary interpretation of their emergence.

A GPU-based algorithm for fast node label learning in large and unbalanced biomolecular networks

BackgroundSeveral problems in network biology and medicine can be cast into a framework where entities are represented through partially labeled networks, and the aim is inferring the labels (usually binary) of the unlabeled part. Connections represent functional or genetic similarity between entities, while the labellings often are highly unbalanced, that is one class is largely under-represented: for instance in the automated protein function prediction (AFP) for most Gene Ontology terms only few proteins are annotated, or in the disease-gene prioritization problem only few genes are actually known to be involved in the etiology of a given disease. Imbalance-aware approaches to accurately predict node labels in biological networks are thereby required. Furthermore, such methods must be scalable, since input data can be large-sized as, for instance, in the context of multi-species protein networks.ResultsWe propose a novel semi-supervised parallel enhancement of COSNet, an imbalance-aware algorithm build on Hopfield neural model recently suggested to solve the AFP problem. By adopting an efficient representation of the graph and assuming a sparse network topology, we empirically show that it can be efficiently applied to networks with millions of nodes. The key strategy to speed up the computations is to partition nodes into independent sets so as to process each set in parallel by exploiting the power of GPU accelerators. This parallel technique ensures the convergence to asymptotically stable attractors, while preserving the asynchronous dynamics of the original model. Detailed experiments on real data and artificial big instances of the problem highlight scalability and efficiency of the proposed method.ConclusionsBy parallelizing COSNet we achieved on average a speed-up of 180x in solving the AFP problem in the S. cerevisiae, Mus musculus and Homo sapiens organisms, while lowering memory requirements. In addition, to show the potential applicability of the method to huge biomolecular networks, we predicted node labels in artificially generated sparse networks involving hundreds of thousands to millions of nodes.

DCO-MAC: A Hybrid MAC Protocol for Data Collection in Underwater Acoustic Sensor Networks.

In underwater acoustic sensor networks (UASNs), medium access control (MAC) is an important issue because of its potentially significant effect on the network performance. However, designing a suitable MAC protocol for the UASN is challenging because of the specific characteristics of the underwater acoustic channel and network, such as limited available bandwidth, long propagation delay, high bit-error-rate, and sparse network topology. In addition, as the traffic load is non-uniformly distributed in a UASN for data collection, it is essential to consider the application feature for the MAC protocol. In this paper, we propose a MAC protocol in a data-collection-oriented UASN, abbreviated as the DCO-MAC protocol. In the proposed protocol, the network is partitioned into two kinds of sub-networks according to the traffic load. A contention-based MAC protocol is used in the sub-network with a light traffic load, while a reservation-based MAC protocol is used in the sub-network with a heavy traffic load. Meanwhile, the DCO-MAC protocol supports the access of mobile nodes. The theoretical analysis and simulation results demonstrate that, in a UASN for data collection, the proposed MAC protocol outperforms the other existing MAC protocols, in terms of the network throughput, end-to-end packet delay, energy overhead, and fairness.

MIMO HetNet IEEE 802.11p–LTE deployment in a vehicular urban environment

Intelligent Transportation Systems (ITS) are receiving significant interest due to a wide variety of applications such as road safety, traffic efficiency and infotainment. To satisfy the diverse vehicular application requirements, this paper presents a detailed comparative performance simulation study over IEEE 802.11p, LTE and the proposed Heterogeneous (HetNet) IEEE 802.11p–LTE network for urban environments. Therefore, a hybrid handover algorithm is proposed, combining the IEEE 802.11p-based multihop clustering and the Long-Term Evolution (LTE) networks, with the goal of achieving a high data packet delivery ratio (PDR) and low delay. The COST-231 Walfisch–Ikegami propagation model is evaluated, incorporating 3D antenna radiation patterns measured in an Anechoic Chamber. The performance evaluation analysis is validated through VEINS–OMNeT++ network simulator over several performance metrics such as Throughput, Packet Delivery Ratio, End-to-End Delay and SNIR Lost Packets for a variety of vehicle density and vehicle speed under both SISO and MIMO open loop 2×2 Spatial Multiplexing (SM) scheme antenna techniques respectively. The results indicate that the HetNet 802.11p–LTE network is the most prominent solution in urban environments outperforming both IEEE 802.11p and LTE in terms of delay, reliability, and scalability, whereas LTE offers acceptable performance for sparse network topologies.

LTE and IEEE 802.11p for vehicular networking: a performance evaluation

Various wireless communication systems exist, which enable a wide range of applications and use cases in the vehicular environment. These applications can be grouped into three types, namely, road safety, traffic efficiency, and infotainment, each with its own set of functional and performance requirements. In pursuance of assisting drivers to travel safely and comfortably, several of these requirements have to be met simultaneously. While the coexistence of multiple radio access technologies brings immense opportunities towards meeting most of the vehicular networking application requirements, it is equally important and challenging to identify the strength and weaknesses of each technology and understand which technology is more suitable for the given networking scenario. In this paper, we evaluate two of the most viable communication standards, Institute of Electrical and Electronics Engineers (IEEE) 802.11p and long-term evolution (LTE) by 3rd Generation Partnership Project for vehicular networking. A detailed performance evaluation study of the standards is given for a variety of parameter settings such as beacon transmission frequency, vehicle density, and vehicle average speed. Both standards are compared in terms of delay, reliability, scalability, and mobility support in the context of various application requirements. Furthermore, through extensive simulation-based study, we validated the effectiveness of both standards to handle different application requirements and share insight for further research directions. The results indicate that IEEE 802.11p offers acceptable performance for sparse network topologies with limited mobility support. On the other hand, LTE meets most of the application requirements in terms of reliability, scalability, and mobility support; however, it is challenging to obtain stringent delay requirements in the presence of higher cellular network traffic load.

Joint Network Optimization and Downlink Beamforming for CoMP Transmissions Using Mixed Integer Conic Programming

Coordinated multipoint (CoMP) transmission is a promising technique to mitigate intercell interference and to increase system throughput in single-frequency reuse networks. Despite the remarkable benefits, the associated operational costs for exchanging user data and control information between multiple cooperating base stations (BSs) limit practical applications of CoMP processing. To facilitate wide usage of CoMP transmission, we consider in this paper the problem of joint network optimization and downlink beamforming (JNOB), with the objective to minimize the overall BS power consumption (including the operational costs of CoMP transmission) while guaranteeing the quality-of-service (QoS) requirements of the mobile stations (MSs). We address this problem using a mixed integer second-order cone program (MI-SOCP) framework and develop an extended MI-SOCP formulation that admits tighter continuous relaxations, which is essential for reducing the computational complexity of the branch-and-cut (BnC) method. Analytic studies of the MI-SOCP formulations are carried out. Based on the analyses, we introduce efficient customizing strategies to further speed up the BnC algorithm through generating tight lower bounds of the minimum total BS power consumptions. For practical applications, we develop polynomial-time inflation and deflation procedures to compute high-quality solutions of the JNOB problem. Numerical results show that the inflation and deflation procedures yield total BS power consumptions that are close to the lower bounds, e.g., exceeding the lower bounds by about 12.9% and 9.0%, respectively, for a network with 13 BSs and 25 MSs. Simulation results also show that minimizing the total BS power consumption results in sparse network topologies and reduced operational overhead in CoMP transmission and that some of the BSs are switched off when possible.

An energy-efficient topology construction algorithm for wireless sensor networks

Topology management schemes have emerged as promising approaches for prolonging the lifetime of the wireless sensor networks (WSNs). The connected dominating set (CDS) concept has also emerged as the most popular method for energy-efficient topology control in WSNs. A sparse CDS-based network topology is highly susceptible to partitioning, while a dense CDS leads to excessive energy consumption due to overlapped sensing areas. Therefore, finding an optimal-size CDS with which a good trade-off between the network lifetime and network coverage can be made is a crucial problem in CDS-based topology control. In this paper, a degree-constrained minimum-weight version of the CDS problem, seeking for the load-balanced network topology with the maximum energy, is presented to model the energy-efficient topology control problem in WSNs. A learning automata-based heuristic is proposed for finding a near optimal solution to the proxy equivalent degree-constrained minimum-weight CDS problem in WSN. A strong theorem in presented to show the convergence of the proposed algorithm. Superiority of the proposed topology control algorithm over the prominent existing methods is shown through the simulation experiments in terms of the number of active nodes (network topology size), control message overhead, residual energy level, and network lifetime.

Service Discovery in Mobile Ad Hoc Networks Based on Grid

Service discovery is a crucial operation for the usability of mobile ad hoc networks (MANETs). Traditional solutions to service discovery that were adopted on the Internet are not well suited for MANETs because of their dynamic nature. This paper presents a novel service discovery protocol based on a hierarchical grid architecture in an ad hoc network. The geographical area of a MANET is divided into a 2-D logical hierarchical grid. The proposed scheme registers the information of available services to a specific location along a predefined trajectory. To enhance resource availability and effective discovery, each grid cell selects a directory for caching available services. This paper utilizes the transmitting trajectory to improve the efficiency of registration and discovery. First, the service provider registers a service along the proposed register trajectory. The requestor then discovers the service along the discovery trajectory to acquire the service information. This paper also proposes an improved process to avoid the sparse node network topology. This paper adopts analysis and simulation to measure the protocol performance. The analytical and simulation results show that the proposed protocol can raise the discovery success ratio and reduce the discovery cost more than the existing geographical service discovery protocol.

Efficient Topology Control for Ad-Hoc Wireless Networks with Non-Uniform Transmission Ranges

Wireless network topology control has drawn considerable attention recently. Priori arts assumed that the wireless ad hoc networks are modeled by unit disk graphs (UDG), i.e., two mobile hosts can communicate as long as their Euclidean distance is no more than a threshold. However, practically, the networks are never so perfect as unit disk graphs: the transmission ranges may vary due to various reasons such as the device differences, the network control necessity, and the perturbation of the transmission ranges even the transmission ranges are set as the same originally. Thus, we assume that each mobile host has its own transmission range. The networks are modeled by mutual inclusion graphs (MG), where two nodes are connected iff they are within the transmission range of each other. Previously, no method is known for topology control when the networks are modeled as mutual inclusion graphs.The paper proposes the first distributed mechanism to build a sparse power efficient network topology for ad hoc wireless networks with non-uniform transmission ranges. We first extend the Yao structure to build a spanner with a constant length and power stretch factor for mutual inclusion graph. We then propose two efficient localized algorithms to construct connected sparse network topologies. The first structure, called extended Yao-Yao, has node degree at most O(log γ), where γ = maxu maxuv∈MG ru/rv. The second structure, called extended Yao and Sink, has node degree bounded by O(log γ), and is a length and power spanner. The methods are based on a novel partition strategy of the space surrounded each mobile host. Both algorithms have communication cost O(n) under a local broadcasting communication model, where each message has O(log n) bits.

The multi-session bridge

The Multicast BackbONE (MBone) is heavily used to carry Internet video conferences. The low cost and ease-of-use have made the MBone and its associated tools such as sdr, vat, vic, and wb very popular. But HEP video conferences often require many-to-many sparse network topologies, an arrangement that cannot easily take advantage of MBone efficiencies. Moreover, many HEP sites do not support multicast and thus researchers at those sites are cut off from important meetings. In addition, the topology of the MBone can cause poor performance or lack of access for some sites. To solve these problems, HEPNRC has developed the Multi-Session Bridge (msb) and its companion tool vtc. The msb allows an arbitrary number of multicast and unicast sessions to be connected in one or more seamless video conferences. Multiple msb's can also be connected to form a conference virtual network. The msb also extends MBone sessions to non-MBone sites through unicast streams. The msb is being extended to provide secure communications.

Data structures for network algorithms on massively parallel architectures

We develop alternative data structures for the representation of network optimization problems on massively parallel SIMD (i.e. Single Instruction Multiple Data) architectures. These data structures are used to implement a primal-dual algorithm for the optimization of a quadratic function subject to network constraints on the Connection Machine CM-2. In particular, two representations of arbitrary, sparse network topologies - a node-wise and an arc-wise representation - are compared with each other in terms of the efficiency with which certain primal-dual algorithms can be executed. Both implementations are also compared with the performance of the same algorithms when applied to dense, bipartite transportation problems. Particular attention is paid to communication efficiency. To this end, the problem of mapping the problem data to the Connection Machine hypercube network is investigated and a pre-computed, optimized, communication scheme is used. The results, collectively, show that very efficient implementations of network algorithms are possible on massively parallel architectures. This statement is true not only in solving dense, bipartite problems, but, even more importantly, in solving problems with arbitrary, sparse network topologies.