The Iterative Group Implicit (IGI) algorithm is developed for the parallel solution of general structural dynamic problems. In this method the original structure is partitioned into a number of a subdomains. Each subd...
详细信息
The Iterative Group Implicit (IGI) algorithm is developed for the parallel solution of general structural dynamic problems. In this method the original structure is partitioned into a number of a subdomains. Each subdomain is solved independently and therefore concurrently, using any traditional direct solution method. The IGI algorithm is an extension of the Group Implicit (GI) algorithm, and similarly to that method compatibility of the interface degrees of freedom is restored using a mass averaging rule. However, unlike the GI algorithm, in the IGI algorithm an iterative procedure is devised to restore equilibrium at the interface degrees of freedom. The IGI method has the same algorithmic characteristics as the underlying solution method used to solve each subdomain. Furthermore, the solution obtained by this method, once the iteration converges, is the same as the one obtained if the subdomain solution method is used to solve the whole structure. Numerical studies are carried out which demonstrate that the performance of the IGI algorithm is superior to that of the GI algorithm both in terms of accuracy and efficiency. Finally, the IGI method is highly modular and scalable, and therefore very well suited for distributed and parallel computing. Copyright (C) 2000 John Wiley & Sons, Ltd.
The purpose of compact routing is to provide a labeling of the nodes of a network and a way to encode the routing tables, so that routing can be performed efficiently (e.g., on shortest paths) whilst keeping the memor...
详细信息
The purpose of compact routing is to provide a labeling of the nodes of a network and a way to encode the routing tables, so that routing can be performed efficiently (e.g., on shortest paths) whilst keeping the memory-space required to store the routing tables as small as possible, In this paper, we answer a long-standing conjecture by showing that compact routing may also assist in the performance of distributed computations. In particular, we show that a network supporting a shortest path interval routing scheme allows broadcasting, with a message-complexity of O(n), where n is the number of nodes of the network. As a consequence, we prove that O(n) messages suffice to solve leader-election for any graph labeled by a shortest path interval routing scheme, improving the previous known bound of O(m + n). A general consequence of our result is that a shortest path interval routing scheme contains ample structural information to avoid developing ad-hoe or network-specific solutions for basic problems that distributed systems must handle repeatedly.
Focuses on the programming and instrumentation environment for parallel processing in the United States. Use of the increased computing power; Creation of parallel applications with rudimentary environment; Amplificat...
详细信息
Focuses on the programming and instrumentation environment for parallel processing in the United States. Use of the increased computing power; Creation of parallel applications with rudimentary environment; Amplification of techniques from sequential programming environment.
The number of devices, from smartphones to IoT hardware, interconnected via the Internet is growing all the time. These devices produce a large amount of data that cannot be analyzed in any data center or stored in th...
详细信息
The number of devices, from smartphones to IoT hardware, interconnected via the Internet is growing all the time. These devices produce a large amount of data that cannot be analyzed in any data center or stored in the cloud, and it might be private or sensitive, thus precluding existing classic approaches. However, studying these data and gaining insights from them is still of great relevance to science and society. Recently, two related paradigms try to address the above problems. On the one hand, edge computing (EC) suggests to increase processing on edge devices. On the other hand, federated learning (FL) deals with training a shared machine learning (ML) model in a distributed (non-centralized) manner while keeping private data locally on edge devices. The combination of both is known as federated edge learning (FEEL). In this work, we propose an algorithm for FEEL that adapts to asynchronous clients joining and leaving the computation. Our research focuses on adapting the learning when the number of volunteers is low and may even drop to zero. We propose, implement, and evaluate a new software platform for this purpose. We then evaluate its results on problems relevant to FEEL. The proposed decentralized and adaptive system architecture for asynchronous learning allows volunteer users to yield their device resources and local data to train a shared ML model. The platform dynamically self-adapts to variations in the number of collaborating heterogeneous devices due to unexpected disconnections (i.e., volunteers can join and leave at any time). Thus, we conduct comprehensive empirical analysis in a static configuration and highly dynamic and changing scenarios. The public open-source platform enables interoperability between volunteers connected using web browsers and Python processes. We show that our platform adapts well to the changing environment getting a numerical accuracy similar to today's configurations using a given number of homogeneous (hardware and
Recent advances in artificial intelligence have driven increasing intelligent applications at the network edge, such as smart home, smart factory, and smart city. To deploy computationally intensive Deep Neural Networ...
详细信息
Recent advances in artificial intelligence have driven increasing intelligent applications at the network edge, such as smart home, smart factory, and smart city. To deploy computationally intensive Deep Neural Networks (DNNs) on resource-constrained edge devices, traditional approaches have relied on either offloading workload to the remote cloud or optimizing computation at the end device locally. However, the cloud-assisted approaches suffer from the unreliable and delay-significant wide-area network, and the local computing approaches are limited by the constrained computing capability. Towards high-performance edge intelligence, the cooperative execution mechanism offers a new paradigm, which has attracted growing research interest recently. In this paper, we propose CoEdge, a distributed DNN computing system that orchestrates cooperative DNN inference over heterogeneous edge devices. CoEdge utilizes available computation and communication resources at the edge and dynamically partitions the DNN inference workload adaptive to devices' computing capabilities and network conditions. Experimental evaluations based on a realistic prototype show that CoEdge outperforms status-quo approaches in saving energy with close inference latency, achieving up to 25.5% similar to 66.9% energy reduction for four widely-adopted CNN models.
The mobility and ubiquitous access afforded by wireless local area networks (WLANs) and high-performance portable products promise to revolutionize the way we live, work, and play. However, sustained improvements in t...
详细信息
The mobility and ubiquitous access afforded by wireless local area networks (WLANs) and high-performance portable products promise to revolutionize the way we live, work, and play. However, sustained improvements in the throughput of WLANs, while also supporting robust long-range operation, will require the use of multiple antennas at both the mobile terminal and the access point. This article reviews the various space-time coding and decoding technologies employed for capitalizing on the increased capacity of the multiple-input multiple-output (MIMO) radio channel. Also described is a channel sounding campaign performed in office environments, used to scope expected performance of these space-time codes in realistic deployments.
Mobile agent-based middleware shows promise for providing an advanced infrastructure that integrates support protocols, mechanisms, and tools to permit communication and coordination of mobile entities.
Mobile agent-based middleware shows promise for providing an advanced infrastructure that integrates support protocols, mechanisms, and tools to permit communication and coordination of mobile entities.
It is well known that the average case deterministic communication complexity is bounded below by an entropic quantity, which one would now call deterministic information complexity. In this paper we show a correspond...
详细信息
It is well known that the average case deterministic communication complexity is bounded below by an entropic quantity, which one would now call deterministic information complexity. In this paper we show a corresponding upper bound. We also improve known lower bounds for the public coin Las Vegas communication complexity by a constant factor. (c) 2006 Elsevier B.V. All rights reserved.
The Media Accelerating Peer Services system extends P2P infrastructures to improve multimedia services across heterogeneous computing platforms. In this article, we present an architecture and resource management and ...
详细信息
The Media Accelerating Peer Services system extends P2P infrastructures to improve multimedia services across heterogeneous computing platforms. In this article, we present an architecture and resource management and adaptation framework that transcends existing infrastructures to accommodate and accelerate multimedia peer applications and services. We also propose key technology components that support seamless adaptation of resources to enhance quality of service and the building of better tools and applications that utilize the peer-computing network's underlying power
Modern scientific data mainly consist of huge data sets gathered by a very large number of techniques and stored in much diversified and often incompatible data repositories. More in general, in the e-science environm...
详细信息
Modern scientific data mainly consist of huge data sets gathered by a very large number of techniques and stored in much diversified and often incompatible data repositories. More in general, in the e-science environment, it is considered as a critical and urgent requirement to integrate services across distributed, heterogeneous, dynamic "virtual organizations" formed by different resources within a single enterprise. In the last decade, Astronomy has become an immensely data-rich field due to the evolution of detectors (plates to digital to mosaics), telescopes and space instruments. The Virtual Observatory approach consists of the federation under common standards of all astronomical archives available worldwide, as well as data analysis, data mining and data exploration applications. The main drive behind such an effort is that once the infrastructure is complete, it will allow a new type of multi-wavelength, multi-epoch science, which can only be barely imagined. Data mining, or knowledge discovery in databases, while being the main methodology to extract the scientific information contained in such Massive Data Sets (MDS), poses crucial problems since it has to orchestrate complex problems posed by transparent access to different computing environments, scalability of algorithms, reusability of resources, etc. In the present paper we summarize the present status of the MDS in the Virtual Observatory and what is currently done and planned to bring advanced data mining methodologies in the case of the DAME (DAta Mining and Exploration) project. (C) 2010 Elsevier B.V. All rights reserved.
暂无评论