Backbone Network Design Research Articles

P2T: Pyramid Pooling Transformer for Scene Understanding.

Recently, the vision transformer has achieved great success by pushing the state-of-the-art of various vision tasks. One of the most challenging problems in the vision transformer is that the large sequence length of image tokens leads to high computational cost (quadratic complexity). A popular solution to this problem is to use a single pooling operation to reduce the sequence length. This paper considers how to improve existing vision transformers, where the pooled feature extracted by a single pooling operation seems less powerful. To this end, we note that pyramid pooling has been demonstrated to be effective in various vision tasks owing to its powerful ability in context abstraction. However, pyramid pooling has not been explored in backbone network design. To bridge this gap, we propose to adapt pyramid pooling to Multi-Head Self-Attention (MHSA) in the vision transformer, simultaneously reducing the sequence length and capturing powerful contextual features. Plugged with our pooling-based MHSA, we build a universal vision transformer backbone, dubbed Pyramid Pooling Transformer (P2T). Extensive experiments demonstrate that, when applied P2T as the backbone network, it shows substantial superiority in various vision tasks such as image classification, semantic segmentation, object detection, and instance segmentation, compared to previous CNN- and transformer-based networks. The code will be released at https://github.com/yuhuan-wu/P2T.

Research on Defect Identification Method of Engine Lining

The engine lining can prevent fuel from debonding with the insulating layer, isolate heat, prevent combustion and buffer stress. Although its proportion is small, the bonding performance of the lining is directly related to whether the engine can maintain a complete structure, and also determines the stability of fuel combustion. How to identify the defects of the formed lining and complete the comprehensive detection of the lining is of great significance to maintain the stability of the engine structure. In this study, an image acquisition system composed of industrial cameras, line-array lenses and line-light sources is used to achieve the integrity acquisition of the lining surface image. Then, through self-made small-scale data sets for the lining image, the backbone network design, and the improvement of existing detectors and network components, the high-precision identification of image defects is achieved.

Key physical topology features for optical backbone networks via a multilayer correlation analysis

A communication network is a multilayer network comprising various layered technologies, and the underlying physical topology is an important aspect that determines the upper bound of overall system performance, including total communication capacity, cost, and robustness. We expect that understanding the impact of the physical topology on system performance will lead to better optical communication network design in the future, and we thus focus on clarifying the relationship between physical topology features and system performance. There have been various studies on the relationship between topology features and overall network performance. For example, the average number of hops and the cluster coefficient are well known to change network properties in complex networks. From the perspective of optical communication networks, it is known that the algebraic connectivity and average path length are related to total network capacity, and these findings have been applied in physical topology design models. On the other hand, there have been no quantitative comparisons among various topology features, and the comprehensiveness of the population from which these features are extracted is insufficient. Hence, we have developed a multilayer (physical topology and layer 1) correlation analysis framework that enables a quantitative comparison of topology features. We use this framework to numerically examine the relationships between physical topology features and the total communication capacity, cost, and robustness of optical communication networks, including graph features (especially graph spectra) that have not been investigated. The results show that the Laplacian spectral radius and geodesic distance Laplacian spectral radius are strongly related to system performance, in addition to the conventional average number of hops, cluster coefficient, algebraic connectivity, and average path length. We confirm that these correlations hold for the different network sizes and spatial nonuniformity of real optical backbone networks in different countries and regions. The results show that the average path length and cluster coefficient, or the Laplacian spectral radius and geodesic distance Laplacian spectral radius, are important guidelines for physical topology design of real optical backbone networks.

SemiCurv: Semi-Supervised Curvilinear Structure Segmentation.

Recent work on curvilinear structure segmentation has mostly focused on backbone network design and loss engineering. The challenge of collecting labelled data, an expensive and labor intensive process, has been overlooked. While labelled data is expensive to obtain, unlabelled data is often readily available. In this work, we propose SemiCurv, a semi-supervised learning (SSL) framework for curvilinear structure segmentation that is able to utilize such unlabelled data to reduce the labelling burden. Our framework addresses two key challenges in formulating curvilinear segmentation in a semi-supervised manner. First, to fully exploit the power of consistency based SSL, we introduce a geometric transformation as strong data augmentation and then align segmentation predictions via a differentiable inverse transformation to enable the computation of pixel-wise consistency. Second, the traditional mean square error (MSE) on unlabelled data is prone to collapsed predictions and this issue exacerbates with severe class imbalance (significantly more background pixels). We propose a N-pair consistency loss to avoid trivial predictions on unlabelled data. We evaluate SemiCurv on six curvilinear segmentation datasets, and find that with no more than 5% of the labelled data, it achieves close to 95% of the performance relative to its fully supervised counterpart.

Optical Cable Backbone Network Design In The Province Kalimantan South Use Technology DWDM OTN

In designing the backbone network for the province of East Kalimantan using Dense Wavelength Division Multiplexing (DWDM) and Optical Transport Network (OTN) devices which cover 10 districts and cities consisting of 34 segments, of which 34 segments consist of 10 OTN DWDM sites and 16 Optical Land Amplifier sites ( OLA). From the results of the design and simulation of the backbone network, it was found that the value of the Optical Link Power Budget (OLPB) < 50 dBm where this value is the ideal value in deploying a DWDM OTN backbone network, and the range of the Optical Signal Noise Ratio (OSNR) is between 20 dBm < OSNR < 40 dBm, which is the ideal value for OSNR, and has an Optical Rise Time Budget (ORTB) < 60 ps, and also has a segment margin system of > 7 dB.

Optical Cable Backbone Network Design In The Province Kalimantan North Use Technology DWDM OTN

In designing the backbone network for the province of East Kalimantan using Dense Wavelength Division Multiplexing (DWDM) and Optical Transport Network (OTN) devices which cover 10 districts and cities consisting of 34 segments, of which 34 segments consist of 10 OTN DWDM sites and 16 Optical Land Amplifier sites ( OLA). From the results of the design and simulation of the backbone network, it was found that the value of the Optical Link Power Budget (OLPB) < 50 dBm where this value is the ideal value in deploying a DWDM OTN backbone network, and the range of the Optical Signal Noise Ratio (OSNR) is between 20 dBm < OSNR < 40 dBm, which is the ideal value for OSNR, and has an Optical Rise Time Budget (ORTB) < 60 ps, and also has a segment margin system of > 7 dB.

EfficientPose: Efficient human pose estimation with neural architecture search

Human pose estimation from image and video is a key task in many multimedia applications. Previous methods achieve great performance but rarely take efficiency into consideration, which makes it difficult to implement the networks on lightweight devices. Nowadays, real-time multimedia applications call for more efficient models for better interaction. Moreover, most deep neural networks for pose estimation directly reuse networks designed for image classification as the backbone, which are not optimized for the pose estimation task. In this paper, we propose an efficient framework for human pose estimation with two parts, an efficient backbone and an efficient head. By implementing a differentiable neural architecture search method, we customize the backbone network design for pose estimation, and reduce computational cost with negligible accuracy degradation. For the efficient head, we slim the transposed convolutions and propose a spatial information correction module to promote the performance of the final prediction. In experiments, we evaluate our networks on the MPII and COCO datasets. Our smallest model requires only 0.65 GFLOPs with 88.1% PCKh@0.5 on MPII and our large model needs only 2 GFLOPs while its accuracy is competitive with the state-of-the-art large model, HRNet, which takes 9.5 GFLOPs.

Optical Cable Backbone Network Design in West Kalimantan Province Using OTN DWDM Technology

In designing the backbone network for the province of West Kalimantan using Dense Wavelength Division Multiplexing (DWDM) and Optical Transport Network (OTN) devices which cover 14 districts and cities consisting of 30 segments, of which 30 segments consist of 14 DWDM OTN sites and 8 Optical Land Amplifier sites ( OLA). From the results of the design and simulation of the backbone network, it was found that the value of the Optical Link Power Budget (OLPB) < 50 dBm where this value is the ideal value in deploying a DWDM OTN backbone network, and the range of the Optical Signal Noise Ratio (OSNR) is between 20 dBm < OSNR < 40 dBm, which is the ideal value for OSNR, and has an Optical Rise Time Budget (ORTB) < 60 ps, and also has a segment margin system of > 7 dB.

Backbone Network Design of Multi-target Automotive Ethernet

Backbone Network Design of Multi-target Automotive Ethernet

Impact of Margins Evolution Along Ageing in Elastic Optical Networks

We compare the cost efficiency of optical networks based on elastic transponders when accounting for network ageing that impacts link margins. Network design of Germany50, France43, and India71 WDM backbone networks is simulated for providing results during a ten-year period, given different throughputs, traffic growth rates and cost erosions of elastic optical transponders. We also investigate the distribution of the ageing margins and modulation schemes that might be changed during the life of the network, before its end-of-life. We evaluate the potential benefits of finer granularity rate-adaptive transponders to better progressively fit ageing of the margin in wavelength division multiplexing networks. The underlying technology and cost savings are presented for the three aforementioned network core topologies.

Earthquake Tolerant Energy Aware Algorithms: A New Approach to the Design of WDM Backbone Networks

The increasing energy cost of networking facilities, in conjunction with environmental instability and climate change, have led to many research efforts targeting the reduction of the energy consumption in telecommunication networks. Wavelength division multiplexing (WDM) optical networks is the key solution for the continuously growing bandwidth demands of backbone networks. However, the rise of traffic has led to the augmentation of energy consumption in such networks. Moreover, backbone networks in countries with high seismicity suffer from unexpected link failures caused by earthquakes, which in turn, cause a huge amount of bandwidth loss and service disruptions. In this paper, a new power aware algorithm is introduced, which uses actual seismic information in order to maintain tolerant connectivity of the backbone network after a large scale earthquake occurs, while supporting energy efficiency. The objective of this paper is the design of a power and earthquake aware WDM backbone network, by establishing the minimum number of backup links in the virtual topology in order to maintain tolerant connectivity. The proposed scheme is evaluated via extensive simulation results by using actual seismic information provided by the Geodynamic Institute of the National Observatory of Athens, and it is verified that both goals of earthquake tolerance and energy efficiency are achieved.

Integer programming models and branch-and-cut approaches to generalized {0,1,2}-survivable network design problems.

In this article, we introduce the Generalized {0,1,2}-Survivable Network Design Problem ({0,1,2}-GSNDP) which has applications in the design of backbone networks. Different mixed integer linear programming formulations are derived by combining previous results obtained for the related {0,1,2}-GSNDP and Generalized Network Design Problems. An extensive computational study comparing the correspondingly developed branch-and-cut approaches shows clear advantages for two particular variants. Additional insights into individual advantages and disadvantages of the developed algorithms for different instance characteristics are given.

Layered graph models and exact algorithms for the generalized hop-constrained minimum spanning tree problem

This paper studies the generalized hop-constrained minimum spanning tree problem (GHMSTP) which has applications in backbone network design subject to quality-of-service constraints that restrict the maximum number of intermediate routers along each communication path. Different possibilities to model the GHMSTP as an integer linear program and strengthening valid inequalities are studied. The obtained formulations are compared theoretically, i.e., by means of their linear programming relaxation. In addition, branch-and-cut approaches based on these formulations are developed and compared in a computational study.

On the optimization of CAPEX and OPEX for the design of island-based translucent optical backbone networks

The transmission reach of signals in optical transmission systems is limited. To go beyond these transparent reach limits, signal regeneration is necessary to re-amplify, re-shape, and re-time the optical signals. Translucent optical networks use a set of sparsely but strategically placed regenerators for the purpose of signal regeneration. These networks are emerged as a promising solution achieving both opaque-networks Quality-of-Transmission (QoT) requirements and transparent-networks cost-effectiveness. Despite massive progress, there are many outstanding issues regarding the design of translucent networks. Mainly, two approaches have been proposed in the literature to deal with the design of such networks, namely, island-of-transparency and sparse-regeneration. In our present work, we adopted the island-based approach which suits much better the operator׳s expectations especially in terms of network scalability and operational expenses, to address the problem of designing translucent WDM backbone networks. Based on such an approach, we present an original two-phase heuristic which locates the regeneration-sites and determines the number of electronic regenerators to be installed in, according to an efficient strategy that aims at minimizing both Capital Expenditures (CAPEX) and Operational Expenditures (OPEX) inherent to regeneration while ensuring QoT requirements. Compared to previously presented island-based approaches, our algorithm is, to the best of our knowledge, the first island-based algorithm that aims at minimizing simultaneously CAPEX and OPEX while using an accurate QoT estimator incorporating all major linear and nonlinear physical-layer impairment effects during the design and dimensioning process. The performances of the proposed heuristic are evaluated through multiple simulations and compared to those of previously presented approaches in order to highlight the main improvements.

Approximation algorithm for optimal location of concentrators and design of Capacitated Survivable Backbone Networks

We focus on the problem of selecting concentrator locations to minimise the access network costs. This falls under the general class of Uncapacitated Facility Location (UFL) problem. We present an approximation algorithm along with some computational results. We also study the problem of designing a Capacitated Survivable Backbone Network (CSBN) with the property that there is at least one route present between all the communicating node pairs in the case of a single link failure. Mathematical formulation to efficiently solve the survivable backbone network design problem is presented. The methodologies developed are applied to a real world setting obtaining optimal or near-optimal results.

The generalized minimum edge‐biconnected network problem: Efficient neighborhood structures for variable neighborhood search

AbstractWe consider the generalized minimum edge‐biconnected network problem where the nodes of a graph are partitioned into clusters and exactly one node from each cluster is required to be connected in an edge‐biconnected way. Instances of this problem appear, for example, in the design of survivable backbone networks. We present different variants of a variable neighborhood search approach that utilize different types of neighborhood structures, each of them addressing particular properties as spanned nodes and/or the edges between them. For the more complex neighborhood structures, we apply efficient techniques—such as a graph reduction—to essentially speed up the search process. For comparison purposes, we use a mixed integer linear programming formulation based on multi‐commodity flows to solve smaller instances of this problem to proven optimality. Experiments on such instances indicate that the variable neighborhood search is also able to identify optimal solutions in the majority of test runs, but within substantially less time. Tests on larger Euclidean and random instances with up to 1,280 nodes, which could not be solved to optimality by mixed integer programming, further document the efficiency of the variable neighborhood search. In particular, all proposed neighborhood structures are shown to contribute significantly to the search process. © 2010 Wiley Periodicals, Inc. NETWORKS, 2010

A quadratic optimization method for connectivity and coverage control in backbone-based wireless networks

The use of directional wireless communications to form flexible mesh backbone networks, which provide broadband connectivity to capacity-limited wireless networks or hosts, promises to circumvent the scalability limitations of traditional homogeneous wireless networks. The main challenge in the design of directional wireless backbone (DWB) networks is to assure backbone network requirements such as coverage and connectivity in a dynamic wireless environment. This paper considers the use of mobility control, as the dynamic reposition of backbone nodes, to provide assured coverage-connectivity in dynamic environments. This paper presents a novel approach to the joint coverage-connectivity optimization problem by formulating it as a quadratic minimization problem. Quadratic cost functions for network coverage and backbone connectivity are defined in terms of the square distance between neighbor nodes, which are related to the actual energy usage of the network system. Our formulation allows the design of self-organized network systems which autonomously achieve energy minimizing configurations driven by local forces exerted on network nodes. The net force on a backbone node is defined as the negative energy gradient at the location of the backbone node. A completely distributed algorithm is presented that allows backbone nodes to adjust their positions based on information about neighbors’ position only. We present initial simulation results that show the effectiveness of our force-based mobility control algorithm to provide network configurations that optimize both network coverage and backbone connectivity in different scenarios. Our algorithm is shown to be adaptive, scalable and self-organized.

Minimum concave cost multicommodity network design

The minimum concave cost multicommodity network design problem (MCMNDP) arises in many application areas, such as transportation planning, energy distribution systems and especially in the design of both packet and circuit switching backbone networks. Exact concave cost optimization algorithms have been developed but they are applicable only if the network size is small. Therefore, MCMNDP is usually solved using non-exact iterative methods. In this paper, such heuristic techniques proposed within the context of circuit switching and packet switching network design are evaluated in detail. Following a comprehensive literature survey, Yaged’s linearization, Minoux’s greedy and Minoux’s accelerated greedy methods have been selected for the circuit switching network design case for further investigation. Minoux’s greedy methods are found to create routes that include cycles causing degradation in the quality of the solution; therefore, we propose a simple but effective modification scheme as a cycle elimination strategy. Similarly, but within the context of packet switching network design, Gerla and Kleinrock’s concave branch elimination, Gersht’s greedy, and Stacey’s concave link elimination methods have been selected for further investigation. All of these methods consider aggregate flows on each link, simultaneously re-routing more than one commodity in one step. In this paper, we propose an alternative disaggregate approach, where only one commodity is handled at a time. Our final proposal is the adaptation of the algorithms proposed for circuit switching network design to the packet switching case. Then an extensive comparative computational study is performed for a number of networks and cost structures to help establish the best method with respect to time and solution quality. Our computational results have shown that the performances of the MCMNDP algorithms heavily depend on the network type and the cost structure. The results have also revealed that our proposed modification to Minoux greedy to eliminate cycles leads to considerable improvements and our proposed disaggregate approach gives the best result in some networks with certain cost structures.

Applications of genetic algorithms to optimal multilevel design of MPLS-based networks

This paper proposes a design methodology based on the application of genetic algorithms (GA) to find a minimal-cost topological structure of MPLS-based networks. MPLS technology is currently deployed in designing the backbone infrastructure of service provider networks whereas other parts of the network are still operated using the traditional IP protocol. This makes the overall topological structure of MPLS-based networks naturally breaks into two prime sub-problems: access network design and backbone network design. The ultimate goal is to identify the locations of label-edge routers and label-switching routers, and to determine the interconnection links and their capacities to accommodate expected traffic demands. The locations of label edge routers depend on the demands of a given set of terminal networks which in turn affect the design of the backbone network. This problem is a highly constrained NP-hard optimization problem for which exact solution approaches do not scale well. We first present a multilevel design model that divides the optimal topology design into a set of linear programs. Then, we propose GA-based meta-heuristics for solving them. We also discuss the impact of encoding methods and genetic operators and parameters on the performance. Numerical results for the considered cases show that the proposed methodology is effective and gives optimal or close to optimal solutions as compared with the exact branch and bound method.

IP over optical cross-connect architectures [Topics in Optical Communications

IP over optical network architectures have been extensively discussed within the research literature over the past few years. However, although signaling protocols between IP routers and optical cross-connect networks have been standardized, large IP backbones are not typically deployed over optical cross-connect networks with automatic reconfigurability features, such as automatic restoration or dynamic establishment of new IP links. One of the most important criteria in determining whether an IP backbone should be carried over such an optical cross-connect network is economic viability. In this article we analyze and explore four architectures for a typical large ISP backbone. In contrast with some other published claims, our results suggest it is more economically attractive to bypass an intermediate cross-connect network, given current equipment and IP backbone network design requirements. However, for ISPs who also provide a large volume of private line services, we propose an integrated architecture for IP over optical cross-connect networks that may provide an attractive alternative for providing rapid and cost-effective restoration from network failures