The genetic algorithm paradigm is a well-known heuristic for solving many problems in science and engineering in which candidate solutions, or “individuals”, are manipulated in ways analogous to biological evolution...
详细信息
The genetic algorithm paradigm is a well-known heuristic for solving many problems in science and engineering in which candidate solutions, or “individuals”, are manipulated in ways analogous to biological evolution, to produce new solutions until one with the desired quality is found. As problem sizes increase, a natural question is how to exploit advances in distributed and parallel computing to speed up the execution of genetic algorithms.
This thesis proposes a new distributed architecture for genetic algorithms, based on distributed storage of the individuals in a persistent pool. Processors extract individuals from the pool in order to perform the computations and then insert the resulting individuals back into the pool. Unlike previously proposed approaches, the new approach is tailored for distributed systems in which processors are loosely coupled, failure-prone and can run at different speeds. Proof-of-concept simulation results are presented for four benchmark functions and for a real-world Product Lifecycle Design problem. We have experimented with both the crash failure model and the Byzantine failure model. The results indicate that the approach can deliver improved performance due to the distribution and tolerates a large fraction of processor failures subject to both models.
Self-Reconfigurable Modular Robots typically consist of high number of modules with uniform docking interfaces, allowing them to transform into various shape. Recognizing the shape of such a system composed of hundred...
详细信息
ISBN:
(数字)9798331519957
ISBN:
(纸本)9798331519964
Self-Reconfigurable Modular Robots typically consist of high number of modules with uniform docking interfaces, allowing them to transform into various shape. Recognizing the shape of such a system composed of hundreds of modules is a significant challenge. Given a new configuration, a modular robots system must be able to determine and update its shape dynamically and in a distributed manner. In a previous work, we developed an algorithm that identifies overlapping boxes to cover the entire robot configuration through message-passing, enabling robots to determine a representation of their current shape. However, this algorithm was static and did not react to changes in real time. In this paper, we introduce an updated shape recognition algorithm that dynamically and in real-time recognizes the addition of modules to update the shape description of the entire configuration using local information. The dynamic algorithm to update the shape description is tested in a simulated environment and compared to re-executing the shape recognition algorithm on the whole configuration. The results show the efficiency of our algorithm in updating the robot’s current shape in real time.
In this paper, we consider the all best swap edges problem in a distributed environment. We are given a 2-edge connected positively weighted network X, where all communication is routed through a rooted spanning tree ...
详细信息
In this paper, we consider the all best swap edges problem in a distributed environment. We are given a 2-edge connected positively weighted network X, where all communication is routed through a rooted spanning tree T of X. If a tree edge e = {x, y} fails, the communication network will be disconnected. However, since X is 2-edge connected, communication can be restored by replacing e by non-tree edge e', called a swap edge of e, whose ends lie in different components of T \ {e}. Of all possible swap edges of e, we would like to choose the best, according to four different objective functions. Overall, the problem is to identify the best swap edge for every tree edge, so that in case of any edge failure, the best swap edge can be activated quickly. There are solutions to this problem for a number of cases in the literature. A major concern for all these solutions is to minimize the number of messages. However, especially in fault-transient environments, time is a crucial factor. In this paper we present a novel technique that addresses this problem from a time perspective;in fact, we present a distributed solution that works in linear time with respect to the height h of T for a number of different criteria, while retaining the optimal number of messages and O (delta(x)) space per each processor x of degree delta(x). To the best of our knowledge, there is no prior algorithm for the all best swap edges problem whose asymptotic complexity matches ours in all three measures: time, space, and number of messages. (C) 2020 Elsevier B.V. All rights reserved.
With the advance in mobile network-based systems, dynamic system has become one of the hotspots in fundamental study of distributed systems. In this article, we consider the dynamic system with frequent topology chang...
详细信息
With the advance in mobile network-based systems, dynamic system has become one of the hotspots in fundamental study of distributed systems. In this article, we consider the dynamic system with frequent topology changes arising from node mobility or other reasons, which is also referred to as dynamic network. With the model of dynamic network, fundamental distributed computing problems, such as information dissemination and election, can be formally studied with rigorous correctness. Our work focuses on the node counting problem in dynamic environments. We first define two new dynamicity models, named (Q, S)-distance and (Q, S)*-distance, which describe dynamic changes of information propagation time against topology changes. Based on these two models, we design three different counting algorithms which basically adopt the approach of diffusing computation. These algorithms mainly differ in communication cost due to different information collection procedures. The correctness of all the algorithms is formally proved and their performance is evaluated via both theoretical analysis and experimental simulations.
Graph coloring is often used in parallelizing scientific computations that run in distributed and multi-GPU environments;it identifies sets of independent data that can be updated in parallel. Many algorithms exist fo...
详细信息
Graph coloring is often used in parallelizing scientific computations that run in distributed and multi-GPU environments;it identifies sets of independent data that can be updated in parallel. Many algorithms exist for graph coloring on a single GPU or in distributed memory, but to the best of our knowledge, hybrid MPI+GPU algorithms have been unexplored until this work. We present several MPI+GPU coloring approaches based on the distributed coloring algorithms of Gebremedhin et al. and the shared-memory algorithms of Deveci et al. The on-node parallel coloring uses implementations in KokkosKernels, which provide parallelization for both multicore CPUs and GPUs. We further extend our approaches to compute distance-2 and partial distance-2 colorings, giving the first known distributed, multi-GPU algorithm for these problems. In addition, we propose a novel heuristic to reduce communication for recoloring in distributed graph coloring. Our experiments show that our approaches operate efficiently on inputs too large to fit on a single GPU and scale up to graphs with 76.7 billion edges running on 128 GPUs.
The author's dissertation, entitled "Scalable SAT Solving and its Application", advances the efficient resolution of instances of the propositional satisfiability (SAT) problem, one of the prototypical &...
详细信息
Synchronous and asynchronous algorithms are presented for distributed minimax optimization. The objective here is to realize the minimization of the maximum of component functions over the standard multi-agent network...
详细信息
Synchronous and asynchronous algorithms are presented for distributed minimax optimization. The objective here is to realize the minimization of the maximum of component functions over the standard multi-agent network, where each node of the network knows its own function and it exchanges its decision variable with its neighbors. In fact, the proposed algorithms are standard consensus and gossip based subgradient methods, while the original minimax optimization is recast as minimization of the sum of component functions by using a p-norm approximation. A scalable step size depending on the approximation ratio p is also presented in order to avoid slow convergence. Numerical examples illustrate that the algorithms with this step size work well even in the high approximation ratios.
Determining the network size is a critical process in numerous areas (e.g., computer science, logistic, epidemiology, social networking services, mathematical modeling, demography, etc.). However, many modern real-wor...
详细信息
Determining the network size is a critical process in numerous areas (e.g., computer science, logistic, epidemiology, social networking services, mathematical modeling, demography, etc.). However, many modern real-world systems are so extensive that measuring their size poses a serious challenge. Therefore, the algorithms for determining/estimating this parameter in an effective manner have been gaining popularity over the past decades. In the paper, we analyze five frequently applied distributed consensus gossip-based algorithms for network size estimation in multi-agent systems (namely, the Randomized gossip algorithm, the Geographic gossip algorithm, the Broadcast gossip algorithm, the Push-Sum protocol, and the Push-Pull protocol). We examine the performance of the mentioned algorithms with bounded execution over random geometric graphs by applying two metrics: the number of sent messages required for consensus achievement and the estimation precision quantified as the median deviation from the real value of the network size. The experimental part consists of two scenarios-the consensus achievement is conditioned by either the values of the inner states or the network size estimates-and, in both scenarios, either the best-connected or the worst-connected agent is chosen as the leader. The goal of this paper is to identify whether all the examined algorithms are applicable to estimating the network size, which algorithm provides the best performance, how the leader selection can affect the performance of the algorithms, and how to most effectively configure the applied stopping criterion.
It is shown that the termination detection problem for distributed computations can be modeled as an instance of the garbage collection problem. Consequently, algorithms for the termination detection problem are obtai...
详细信息
It is shown that the termination detection problem for distributed computations can be modeled as an instance of the garbage collection problem. Consequently, algorithms for the termination detection problem are obtained by applying transformations to garbage collection algorithms. The transformation can be applied to collectors of the ''mark-and-sweep'' type as well as to reference-counting garbage collectors. As examples, the scheme is used to transform the distributed reference-counting protocol of Lermen and Maurer, the weighted-reference-counting protocol, the local-reference-counting protocol, and Ben-Ari's mark-and-sweep collector into termination detection algorithms. Known termination-detection algorithms as well as new variants are obtained.
In this paper, the optimal variational generalized Nash equilibrium(v-GNE) seeking problem in merely monotone games with linearly coupled cost functions is investigated, in which the feasible strategy domain of each a...
详细信息
In this paper, the optimal variational generalized Nash equilibrium(v-GNE) seeking problem in merely monotone games with linearly coupled cost functions is investigated, in which the feasible strategy domain of each agent is coupled through an affine constraint. A distributed algorithm based on the hybrid steepest descent method is first proposed to seek the optimal v-GNE. Then, an accelerated algorithm with relaxation is proposed and analyzed, which has the potential to further improve the convergence speed to the optimal v-GNE. Some sufficient conditions in both algorithms are obtained to ensure the global convergence towards the optimal v-GNE. To illustrate the performance of the algorithms, numerical simulation is conducted based on a networked Nash-Cournot game with bounded market capacities.
暂无评论