Showing papers on "Network topology published in 1989"

The Cascade-Correlation Learning Architecture

Abstract: Cascade-Correlation is a new architecture and supervised learning algorithm for artificial neural networks. Instead of just adjusting the weights in a network of fixed topology. Cascade-Correlation begins with a minimal network, then automatically trains and adds new hidden units one by one, creating a multi-layer structure. Once a new hidden unit has been added to the network, its input-side weights are frozen. This unit then becomes a permanent feature-detector in the network, available for producing outputs or for creating other, more complex feature detectors. The Cascade-Correlation architecture has several advantages over existing algorithms: it learns very quickly, the network determines its own size and topology, it retains the structures it has built even if the training set changes, and it requires no back-propagation of error signals through the connections of the network.

Stability of analog neural networks with delay

Abstract: Continuous-time analog neural networks with symmetric connections will always converge to fixed points when the neurons have infinitely fast response, but can oscillate when a small time delay is present. Sustained oscillation resulting from time delay is relevant to hardware implementations of neural networks where delay due to the finite switching speed of amplifiers can be appreciable compared to the network relaxation time. We analyze the dynamics of continuous-time analog networks with delay, and show that there is a critical delay above which a symmetrically connected network will oscillate. Two different stability analyses are presented for low and high neuron gain. The results are useful as design criteria for building fast but stable electronic networks. We find that for some connection topologies, a delay much smaller than the relaxation time can lead to oscillation, whereas for other topologies, including associative memory networks, even long delays will not produce oscillation. The most oscillation-prone network configuration is the all-inhibitory network; in this configuration, the critical delay for oscillation is smaller than the network relaxation time by a factor of N, the size of the network. Theoretical results are compared with numerical simulations and with experiments performed on a small (eight neurons) electronic network with controllable delay.

Dynamic Node Creation in Backpropagation Networks

Abstract: This paper introduces a new method called Dynamic Node Creation (DNC) which automatically grows BP networks until the target problem is solved. DNC sequentially adds nodes one at a time to the hidden layer(s) of the network until the desired approximation accuracy is achieved. Simulation results for parity, symmetry, binary addition, and the encoder problem are presented. The procedure was capable of finding known minimal topologies in many cases, and was always within three nodes of the minimum. Computational expense for finding the solutions was comparable to training normal BP networks with the same final topologies. Starting out with fewer nodes than needed to solve the problem actually seems to help find a solution. The method yielded a solution for every problem tried.

A trade-off between space and efficiency for routing tables

Abstract: Two conflicting goals play a crucial role in the design of routing schemes for communication networks. A routing scheme should use paths that are as short as possible for routing messages in the network, while keeping the routing information stored in the processors' local memory as succinct as possible. The efficiency of a routing scheme is measured in terms of its stretch factor-the maximum ratio between the length of a route computed by the scheme and that of a shortest path connecting the same pair of vertices.Most previous work has concentrated on finding good routing schemes (with a small fixed stretch factor) for special classes of network topologies. In this paper the problem for general networks is studied, and the entire range of possible stretch factors is examined. The results exhibit a trade-off between the efficiency of a routing scheme and its space requirements. Almost tight upper and lower bounds for this trade-off are presented. Specifically, it is proved that any routing scheme for general n-vertex networks that achieves a stretch factor k ≥ 1 must use a total of O(n1+1/(2k+4)) bits of routing information in the networks. This lower bound is complemented by a family K(k) of hierarchical routing schemes (for every k ≥ l) for unit-cost general networks, which guarantee a stretch factor of O(k), require storing a total of O(k3n1+(1/h)logn)- bits of routing information in the network, name the vertices with O(log2n)-bit names and use O(logn)-bit headers.

Communications network state and topology monitor

Abstract: A system gathers and displays information concerning status of a communications network without overloading the communications channels in the network. The monitoring system includes a monitor node, including an operator input interface. The monitor node is coupled to a first switching node in the distributed switching nodes of the network. The monitor node includes a first application maintaining topology data indicating the topology of the network and supporting a first protocol for updating the data with the first switching node. In addition, the monitor node includes a second application maintaining a list of alarm conditions entered in the node event logs in the network, and supporting a second protocol for updating the list with the plurality of distributed switching nodes. A third application runs in the monitor node for maintaining a monitor database indicating the configuration of the switching nodes as it is entered in the node configuration databases in the network. This third application also supports a third protocol for updating the monitor database with the plurality of distributed switching nodes for updates to the node configuration database. Distributed applications in the switching nodes support the protocols with the monitor node without interference with communications tasks in the network. A display application that is responsive to operator inputs identifying a subject node or other object in the network having plural windows on a display monitor, presents configuration data about the subject node, the network topology and the alarm conditions to the operator.

Comparison of deflection and store-and-forward techniques in the Manhattan Street and Shuffle-Exchange Networks

Abstract: The Manhattan Street Network (MS-Net) and Shuffle-Exchange Network (SX-Net) are two-connected networks with significantly different topologies. Fixed-size packets are transmitted between nodes in these networks. The nodes are synchronized so that all of the packets that are received by a node within a slot transmission time arrive at a switching point simultaneously. Instead of storing large numbers of packets at intermediate nodes, a deflection strategy similar to hot-potato routing is used. There are characteristics of the MS-Net that make it well suited for deflection routing. With no buffer, 55-70% of the throughput with an infinite number of buffers has been obtained; with a single buffer per node, the throughput increases to 80-90%. With uniform load the throughput does not decrease significantly as the network utilization increases. Therefore, additional flow control mechanisms are not required to achieve the highest network throughput. The SX-Net does not have the above characteristics of the MS-Net. However, deflection routing still provides a significant portion of the available throughput. In the SX-Net, more buffers are required than in the MS-Net, and a flow control mechanism must be used to achieve the greatest throughput. >

A system for routing and capacity assignment in computer communication networks

Abstract: The combined problem of selecting a primary route for each communicating pair and a capacity value for each link in computer communication networks is considered. The network topology and traffic characteristics are given: a set of candidate routes and of candidate capacities for each link are also available. The goal is to obtain the least costly feasible design where the costs include both capacity and queuing components. Lagrangean relaxation and subgradient optimization techniques were used to obtain verifiable solutions to the problem. The method was tested on several topologies, and in all cases good feasible solutions, as well as tight lower bounds, were obtained. The model can be generalized to deal with different classes of customers, characterized by different priorities, message lengths, and/or delay requirements. >

Detection of Topology Errors by State Estimation

Abstract: In modern energy management systems (EMS), there are two types of measurement data collected by the SCADA (supervisory control and data acquisition) system, namely, status data of breakers and switches, and analog data of real and reactive power flows, injections, and bus voltages. The status data are used to determine real-time topology of the network. The analog data are used to determine line and transformer loading and voltage profile. These data are noisy due to measurement errors, communication noise, missing data, etc. In addition to simple checking of the analog data locally, in most modern EMS, state estimation is used to process these data globally to correct errors in the raw analog measurement data. In this paper, the use of global information in state estimation to correct errors in the direct measurements of the status data is proposed and conditions for detectability of errors are analyzed. The telemetered data of breaker and switch status are processed in the EMS computer to determine the present network topology of the system, and this function is called network topology processor. Errors in status data will show up as errors in the network topology. Sasson et al and Dy Liacco et al used a tree search algorithm for the network topology processor. The method is widely used in practice. Bonanomi et al proposed a sequential search method through the network graph. Recently Lugtu et al suggested the approach of using state estimation results for topology error detection.

Dynamic node creation in backpropagation networks

Abstract: Summary form only given. A novel method called dynamic node creation (DNC) that attacks issues of training large networks and of testing networks with different numbers of hidden layer units is presented. DNC sequentially adds nodes one at a time to the hidden layer(s) of the network until the desired approximation accuracy is achieved. Simulation results for parity, symmetry, binary addition, and the encoder problem are presented. The procedure was capable of finding known minimal topologies in many cases, and was always within three nodes of the minimum. Computational expense for finding the solutions was comparable to training normal backpropagation (BP) networks with the same final topologies. Starting out with fewer nodes than needed to solve the problem actually seems to help find a solution. The method yielded a solution for every problem tried. BP applied to the same large networks with randomized initial weights was unable, after repeated attempts, to replicate some minimum solutions found by DNC. >

Detection of topology errors by state estimation (power systems)

Abstract: Errors in the telemetered data of breaker and switch status, through the network topology processor in the EMS (energy management system) computer, may result in errors in the determination of the current network topology of the system. The use of normalized residuals that result from state estimation is proposed for the detection of topology errors. Three types of topology errors are considered: line or transformer outage, bus split, and shunt capacitor/reactor switching. Conditions for detectability of topology errors are presented. The conditions are tested on the IEEE 30 bus system, and the results confirm the theoretical predictions. The problem of topology error identification is also discussed. >

Topology design and bandwidth allocation in ATM nets

Abstract: The design of a P/S network embedded into a backbone facility network is discussed. The problem is formulated as a network optimization problem where a congestion measure based on the average packet delay is minimized, subject to capacity constraints posed by the underlying facility trunks. The variables in this problem are the routing on the express pipes (i.e. the channels that interconnect the P/S modes) and the allocation of bandwidth to such pipes. An efficient algorithm is presented for the solution of the above problem and it is applied to some representative examples. It is shown that for some test cases the congestion measure is substantially reduced with respect to the values obtained when the embedded topology is kept identical to the backbone topology. Dynamic reconfiguration schemes where the embedded topology is periodically adjusted to track the fluctuations in traffic requirements are discussed. >

Hypermedia topologies and user navigation

Abstract: One of the major problems confronting users of large hypermedia systems is that of navigation: knowing where one is, where one wants to go, and how to get there from here. This paper contributes to this problem in three steps. First, it articulates a number of navigational strategies that people use in physical (geographical) navigation. Second, it correlates these with various graph topologies, showing how and why appropriately restricting the connectivity of a hyperbase can improve the ability of users to navigate. Third, it analyzes some common hypermedia navigational mechanisms in terms of navigational strategies and graph topology.

An O(logN) deterministic packet routing scheme

Abstract: We present a deterministic O(log N) time algorithm for the problem of routing an arbitrary permutation on an N-processor bounded-degree network with bounded buffers.Unlike all previous deterministic solutions to this problem, our routing scheme does not reduce the routing problem to sorting and does not use the Ajtai, Komlos and Szemeredi sorting network [AKS]. Consequently, the constant in the run time of our routing scheme is substantially smaller, and the network topology is significantly simpler.

Process synchronization: design and performance evaluation of distributed algorithms

Abstract: The author presents a simple solution for the committee coordination problem, which encompasses the synchronization and exclusion problems associated with implementing multiway rendezvous, and shows how it can be implemented to develop a family of algorithms. The algorithms use message counts to solve the synchronization problem, and they solve the exclusion problem by using a circulating token or by using auxiliary resources as in the solutions for the dining or drinking philosophers' problems. Results of a simulation study of the performance of the algorithms are presented. The experiments measured the response time and message complexity of each algorithm as a function of variations in the model parameters, including network topology and level of conflict in the system. The results show that the response time for algorithms proposed is significantly better than for existing algorithms, whereas the message complexity is considerably worse. >

Fiberoptic circuit network design under reliability constraints

Abstract: A general mathematical model for a network design problem with reliability constraints and a revised formulation which seems particularly appropriate for fiber-optics networks is presented. Upper and lower bounding procedures based on continuous relaxations of this modified formulation are described. Preliminary computational results are reported. Limited computational results indicate a good performance of the algorithm, producing a gap between lower and upper bounds that is sufficiently small for a branch-and-bound procedure to be applicable. >

An optimal shortest-path routing policy for network computers with regular mesh-connected topologies

Abstract: A probabilistic routing policy, the Z/sup 2/ (zigzag) routing policy, is presented within the class of nonadaptive, shortest-path routing policies for regular mesh-connected topologies such as n-dimensional toroids and hypercubes. The focus of the research is routing in networks of computers in a distributed computing environment, where constituent subcomputers are organized in a mesh-connected topology and communication among individual computers takes places by some form of message exchange. The authors prove the optimality of this policy with respect to two criteria: (1) maximizing the probability of reaching the destination from a given source without delays at intermediate nodes; and (2) minimizing the expected lifetime of a message. >

Computer-aided design procedures for survivable fiber optic networks

Abstract: The authors describe methods to design cost-effective survivable telecommunications networks that use fiber-optic transmission links. One of these methods utilizes optical switching devices to implement route diversity during cable cuts. These methods have been incorporated into a software tool consisting of three modules: a topology generator, a circuit to DS3 bundler, and a multiplex layout system. This tool is compact enough to run on a personal computer. Each of these three modules is described, and sample results are provided. >

Passive optical subscriber loops with multiaccess

Abstract: Deployment of optical fiber in the subscriber loop may eventually lead to a full-scale deployment of the broadband network. A network architecture that can be introduced inexpensively to meet the near-term demand and upgraded smoothly to support future needs is desired. The feasibility of applying multiaccess architectures to subscriber loops is studied, and five architectures for passive optical subscriber loops that meet this need are presented. All of these architectures use a double-star topology and dense wavelength division multiplexing in the downstream direction; however, various topologies and multiaccess techniques are used in the upstream direction. Limitations on node size, frame synchronization, cost, privacy and security, and standards are discussed. >

Using the multistage cube network topology in parallel supercomputers

Abstract: A critical component of any large-scale parallel processing system is the interconnection network that provides a means for communication along the system's processors and memories. Attributes of the multistage cube topology that have made it an effective basis for interconnection networks and the subject of much ongoing research are reviewed. These attributes include O(N log/sub 2/N) cost for an N-input/output network, decentralized control, a variety of implementation options, good data-permuting capability to support single-instruction-stream/multiple-data-stream (SIMD) parallelism, good throughput to support multiple-instruction-stream/multiple-data-stream (MIMD) parallelism, and ability to be partitioned into independent subnetworks to support reconfigurable systems. Examples of existing systems that use multistage cube networks are considered. The multistage cube topology can be converted into a single-stage network by associating with each switch in the network a processor (and a memory). Properties of systems that use the multistage cube network in this way are examined. >

Supercube: An optimally fault tolerant network architecture

Abstract: A new class of interconnection network topology is proposed for parallel and distributed processing. The attractive features of this class include (a) the network can be constructed for any number of computing nodes, (b) the network is incrementally expandable, i.e., a new node can easily be added to the existing network, (c) it has good fault-tolerant characteristics (measured by the connectivity of the network graph) and (d) it has small delay characteristics (measured by the diameter of the network graph). The node connectivity of the network is equal to the minimum node degree. In this sense the network is optimally fault-tolerant.

Automatic Sizing of Power/Ground (P/G) Networks in VLSI

Abstract: This paper presents a fast and efficient method for sizing power/ground networks. No restrictions on network topology or number of supplying pads are imposed. Wire widths are calculated such that the weighted area of wire segments is minimized while electromigration and voltage drops constraints are fulfilled. The algorithm proposed here runs 50% faster than the best methods reported for tree type network topologies.

Tree-height minimization in pipelined architectures

Abstract: A method of tree-height minimization for networks of commutative and associative operators is proposed. An algorithm is described for minimizing latency and shimming delays in a synchronous data-flow architecture such as that used in pipelined or bit- or digit-serial computation. The algorithm rearranges operator trees to meet the joint goals, often allowing otherwise impossible scheduling constraints to be met. It may also be applied to word-parallel pipelined architectures to optimize operator trees within pipelined stages. The method is evaluated by testing it on several filter examples for which it finds optimal network topologies and schedules. >

Event driven topology broadcast without sequence numbers

Abstract: An algorithm is presented that allows each node in a computer network to maintain a correct view of the network topology despite link and node failures. Reliability is achieved without transmitting any information other than the operational status of links. Messages are only sent in response to topological changes: periodic retransmission is not required. >

On the complexity of radio communication

Abstract: A radio network is a synchronous network of processors that communicate by transmitting messages to their neighbors. A processor receives a message in a given step if and only if it is silent then and precisely one of its neighbors transmits. This stringent rule poses serious difficulties in performing even the simplest tasks. This is true even under the overly optimistic assumptions of centralized coordination and complete knowledge of the network topology. This paper is concerned with lower and upper bounds for the complexity of realizing various communication primitives for radio networks.Our first result deals with the broadcast operation. We prove the existence of a family of radius-2 networks on n vertices for which any broadcast schedule requires at least O(log2n) rounds of transmissions. This matches an upper bound of O(log2n) rounds for networks of radius 2 proved earlier by Bar-Yehuda, Goldreich and Itai [BGI]. It is worth mentioning that this lower bound holds even under optimal centralized coordination, while the (randomized) algorithm of [BGI] is distributed.We then look at the question of simulating two of the standard message-passing models on a radio network. Both models can easily simulate the radio model with no overhead. In the other direction, we propose and study a primitive called the single-round simulation (SRS), enabling the simulation of a single round of an algorithm designed for the standard message models. We give lower bounds for the length of SRS schedules for both models, and supply constructions or existence proofs for schedules of matching (or almost matching) lengths.Finally we give tight bounds for the length of schedules for computing census functions on a radio network.

Structured systems analysis methodology for design of an ATM network architecture

Abstract: The asynchronous transfer mode (ATM) network design problem is treated as a resource-allocation problem. Traffic control is applied to different entities: packets, bursts, calls, and flows. Each of these entities can be controlled locally, within an exchange, or globally, at the network level. This gives rise to a multistrata architecture, where the resource-allocation stratum is defined both by the layer of traffic flow and the control level. Dedicated and virtual connections can be set up in all strata: for bursts, calls, subscriptions, and flows, at the exchange, subnetwork, and network levels. The open systems interconnection protocols are used to map heterogeneous user entities onto uniform network entities. The critical load is defined, which is where burst and call fluctuations begin to result in long queues and big packet delays. The critical load is used to determine thresholds, at the network and exchange level, for the number of virtual or dedicated connections which can be established in each traffic class. >

Topological analysis of packet networks

Abstract: The author describes a unified approach for the topological analysis of nonhierarchical and hierarchical packet networks. The approach differs from previous approaches in adopting an end-to-end mean delay objective and including a variety of practical routing constraints. These include limits on the number of paths allowed in a route, limits on the number of hops allowed in a path, and constraints due to prevalent virtual circuit implementations. For a broad range of networks, quantitative analysis based on this approach provides new insights into the complex relationships between network topology and routing and delay constraints. It is shown that the sole use of a network average delay criterion often leads to network designs that exhibit poor end-to-end mean delays for some node pairs, and that it is possible to configure networks that meet an end-to-end mean delay objective for every node pair at little or no additional cost. >

The HCN: a versatile interconnection network based on cubes

Abstract: This paper introduces a family of interconnection networks for loosely-coupled multiprocessors called Hierarchical Cubic Networks (HCNs). HCNs use the well-known hypercube network as their basic building block. Using a considerably lower number of links per node, HCNs realize lower network diameters than the hypercube. The performance of several well-known applications on a hypothetical system employing the HCN is identical to their performance on a hypercube. HCNs thus enjoy the same advantages as a hypercube, albeit with considerably simpler interconnections.

Robust design and planning of a worldwide intelligent network

Abstract: The authors describe the structure of the worldwide intelligent network (WIN), describe methods for its design and planning, investigate the adequacy of decentralized control for problem-free worldwide call completion, explore the feasibility of adaptive routing and control concepts, discuss network robustness/reliability objectives, and describe a strategy for achieving these objectives for all cooperating international carriers. Several decentralized adaptive routing policies that are particularly attractive in the WIN environment and network performance improvements that can be achieved with the introduction of flexible routing capabilities are characterized. It is shown by example that network integration coupled with flexible routing and bandwidth allocation for preferential treatment of new services provides an effective approach for robust and economical new service provisioning. >

On the communication complexity of generalized 2-D convolution on array processors

Abstract: Several parallel convolution algorithms for array processors with N/sup 2/ processing elements (PEs) connected by mesh, hypercube, and shuffle-exchange topologies, respectively, are presented. The computation time complexity is the same for array processors with different interconnection networks. The communication time complexity, however, varies from network to network, and is the main focus. It is shown that by using inter-PE communication networks efficiently, each PE requires only a small local memory, many unnecessary data transmissions are eliminated, and the overall time complexity (including computation and communication) of algorithms is reduced to O(M/sup 2/). >