distributed inference techniques can be broadly classified into data-distributed and model-distributed schemes. In data-distributed inference (DDI), each worker carries the entire deep neuralnetwork (DNN) model, but ...
详细信息
ISBN:
(纸本)9798350302615
distributed inference techniques can be broadly classified into data-distributed and model-distributed schemes. In data-distributed inference (DDI), each worker carries the entire deep neuralnetwork (DNN) model, but processes only a subset of the data. However, feeding the data to workers results in high communication costs especially when the data is large. An emerging paradigm is model-distributed inference (MDI), where each worker carries only a subset of DNN layers. In MDI, a source device that has data processes a few layers of DNN and sends the output to a neighboring device. This process ends when all layers are processed in a distributed manner. In this paper, we investigate MDI with multiple sources, i.e., when more than one device has data. We design a multi-source MDI (MS-MDI), which optimizes task scheduling decisions across multiple source devices and workers. Experimental results on a real-life testbed of NVIDIA Jetson TX2 edge devices show that MS-MDI improves the inference time significantly as compared to baselines.
Summary form only given, as follows. neuralnetwork models are reviewed from the point of view of parallel distributedprocessing models. More specifically, the interactions between neural models and the computational...
详细信息
ISBN:
(纸本)0780301641
Summary form only given, as follows. neuralnetwork models are reviewed from the point of view of parallel distributedprocessing models. More specifically, the interactions between neural models and the computational model implemented by transputer-based machines are identified so as to define a general strategy for mapping neuralnetworks onto transputers, and therefore achieving the design and a prototype of a neuralnetwork mapper for transputer-based machines.
Y Sky computing is a new computing paradigm leveraging resources of multiple Cloud providers to create a large scale distributed infrastructure. N2Sky is a research initiative promising a framework for the utilization...
详细信息
ISBN:
(纸本)9783030042240;9783030042233
Y Sky computing is a new computing paradigm leveraging resources of multiple Cloud providers to create a large scale distributed infrastructure. N2Sky is a research initiative promising a framework for the utilization of neuralnetworks as services across many Clouds. This involves a number of challenges ranging from the provision, discovery and utilization of services to the management, monitoring, metering and accounting of the infrastructure. Cloud Container technology offers fast deployment, good portability, and high resource efficiency to run large-scale and distributed systems. In recent years, container-based virtualization for applications has gained immense popularity. This paper presents the new N2SkyC system, a framework for the utilization of neuralnetworks as services, aiming for higher flexibility, portability, dynamic orchestration, and performance by fostering microservices and Cloud container technology.
The n-mode fiber optic sensor built has four linearly polarized (LP) modes propagating simultaneously in the fiber, producing a two-dimensional, spatially distributed output intensity pattern. When the fiber is strain...
详细信息
ISBN:
(纸本)0819412015
The n-mode fiber optic sensor built has four linearly polarized (LP) modes propagating simultaneously in the fiber, producing a two-dimensional, spatially distributed output intensity pattern. When the fiber is strained, there is a change in fiber parameters. Oscillating and rotating of the pattern caused by coupling between degenerate modes is observed. Thus the processing of this type of output signal becomes one of a two-dimensional image processor. A neuralnetwork signal processor employing a back propagation algorithm was used in conjunction with the few mode fiber optic sensor to categorize the spatial output patterns from the sensor, thus converting the optical pattern to its corresponding strain value. The testing results show that the neuralnetwork processor is capable of recognizing this kind of image with good accuracy, resulting in strain accuracies within 0.7 percent.
distributed scheduling algorithms for throughput or utility maximization in dense wireless multi-hop networks can have overwhelmingly high overhead, causing increased congestion, energy consumption, radio footprint, a...
详细信息
ISBN:
(纸本)9781665405409
distributed scheduling algorithms for throughput or utility maximization in dense wireless multi-hop networks can have overwhelmingly high overhead, causing increased congestion, energy consumption, radio footprint, and security vulnerability. For wireless networks with dense connectivity, we propose a distributed scheme for link sparsification with graph convolutional networks (GCNs), which can reduce the scheduling overhead while keeping most of the network capacity. In a nutshell, a trainable GCN module generates node embeddings as topology-aware and reusable parameters for a local decision mechanism, based on which a link can withdraw itself from the scheduling contention if it is not likely to win. In mediumsized wireless networks, our proposed sparse scheduler beats classical threshold-based sparsification policies by retaining almost 70% of the total capacity achieved by a distributed greedy max-weight scheduler with 0.4% of the point-to-point message complexity and 2.6% of the average number of interfering neighbors per link.
In this paper, the synchronization characteristics in response to external inputs are investigated for chaotic neuralnetworks with coupled lattices based on the Newman-Watts model. The Newman-Watts model was original...
详细信息
ISBN:
(纸本)9781467314909
In this paper, the synchronization characteristics in response to external inputs are investigated for chaotic neuralnetworks with coupled lattices based on the Newman-Watts model. The Newman-Watts model was originally proposed with a ring-coupled lattice as its initial structure. However, ring-coupled networks are not suitable for a number of applications, including image processing. Therefore, in this paper, the synchronization characteristics in response to external inputs are investigated in a coupled lattice based on a Newman-Watts network. As a result, we find that synchronized clusters are generated in response to spatially distributed external inputs, and recombination of neurons into clusters occurs in the case that the parameter values of a single neuron correspond to those giving chaotic dynamics. Moreover, we explore the possibility that chaotic dynamics are useful for separating two image segments that have similar grayscale values by using the proposed network with synchronization.
Graph neuralnetworks (GNNs) learn representations from network data with naturally distributed architectures, rendering them well-suited candidates for decentralized learning. Oftentimes, this decentralized graph sup...
详细信息
ISBN:
(纸本)9781728176055
Graph neuralnetworks (GNNs) learn representations from network data with naturally distributed architectures, rendering them well-suited candidates for decentralized learning. Oftentimes, this decentralized graph support changes with time due to link failures or topology variations. These changes create a mismatch between the graphs on which GNNs were trained and the ones on which they are tested. Online learning can be used to retrain GNNs at testing time, overcoming this issue. However, most online algorithms are centralized and work on convex problems (which GNNs rarely lead to). This paper proposes the Wide and Deep GNN (WD-GNN), a novel architecture that can be easily updated with distributed online learning mechanisms. The WD-GNN comprises two components: the wide part is a bank of linear graph filters and the deep part is a GNN. At training time, the joint architecture learns a nonlinear representation from data. At testing time, the deep part (nonlinear) is left unchanged, while the wide part is retrained online, leading to a convex problem. We derive convergence guarantees for this online retraining procedure and further propose a decentralized alternative. Experiments on the robot swarm control for flocking corroborate theory and show potential of the proposed architecture for distributed online learning.
In this paper, a fault tolerant CNN (Cellular neuralnetwork) using a small world network connection for image processing in embedded systems is described. In embedded systems, there are problems such as compactness, ...
详细信息
This paper presents a network architecture to interconnect VLSI1 neuralnetwork chips to build a distributed ANN(2) system. The architecture combines techniques from circuit switching and packet switching to provide t...
详细信息
ISBN:
(纸本)9781424435494
This paper presents a network architecture to interconnect VLSI1 neuralnetwork chips to build a distributed ANN(2) system. The architecture combines techniques from circuit switching and packet switching to provide two different service classes: isochronous connections and best-effort packet transfers. The isochronous connections are able to transport the axonal data of artificial neurons between VLSI ANN models that feature a speedup of multiples orders of magnitudes compared to biology. The connections use reserved bandwidth to provide loss-less transmissions as well as a low end-to-end delay with bounded jitter. Best-effort packet transfers use the remaining bandwidth for on-demand multi-purpose communication. The data forwarding is performed between synchronized instances of a dedicated switch architecture used at each network node. The switch is scalable in terms of port numbers and line speed. Its low complexity allows for an implementation within programmable logic or directly within a VLSI neuralnetwork chip. A reference implementation of the proposed network architecture is presented within an existing framework that hosts VLSI neuralnetwork chips operating at speedups of 10(4) to 10(5). The network architecture is further not limited to VLSI neuralnetworks, but it can in principle be used in all network environments that require isochronous connections as well as packet processing.
作者:
Schoonees, J.A.CSIR
Div for Microelectron & Commun Technol Pretoria S Afr
An introduction to artificial neuralnetwork models is presented, along with an overview of their practical application and potential applications in signal processing. Successful neuralnetwork implementations are de...
详细信息
ISBN:
(纸本)0879427094
An introduction to artificial neuralnetwork models is presented, along with an overview of their practical application and potential applications in signal processing. Successful neuralnetwork implementations are described and their performances are compared to those of more traditional signal processing implementations. The Hopfield net, self-organizing feature maps, and the multilayer perceptron are reviewed. Implementation of neural nets in speech synthesis, speech recognition, target identification, image processing, pattern matching, error-correction coding, and neurocomputing are reported. Several ICs currently in production are briefly mentioned.
暂无评论