Hierarchical neural networks based prediction and control of dynamic reconfiguration for multilevel embedded systems

被引：11

作者：

Eddahech, Akram ^{[1
]}

Chtourou, Sofien ^{[2
]}

Chtourou, Mohamed ^{[1
]}

机构：

[1] Ecole Natl Ingn Sfax, Sfax, Tunisia

[2] Inst Super Informat & Multimedia, Sfax 3021, Tunisia

来源：

JOURNAL OF SYSTEMS ARCHITECTURE | 2013年 / 59卷 / 01期

关键词：

Hierarchical prediction; Recurrent neural network; Dynamic reconfiguration; Embedded system; ARCHITECTURES; DESIGN; PERFORMANCE; ALGORITHM;

D O I：

10.1016/j.sysarc.2012.11.002

中图分类号：

TP3 [计算技术、计算机技术];

学科分类号：

0812 ;

摘要：

Multimedia design such as video decoders are typically composed of several communicating tasks. Each task is characterized by its workload variation. The target device of this kind of application contains several processing unit. This calls for a dynamic management of hardware units to improve the QOS of the application and to optimally allocate resources. In this paper, we propose a new architecture based on hierarchical multilevel neural network to model workload variation of each task. The hierarchical structure of this neural network perfectly describes the multilevel decomposition of each hardware unit. The aim of this investigation is to build a design with a control unit that manages the architecture and resource allocation according to the neural network workload prediction. (c) 2012 Elsevier B.V. All rights reserved.

引用

页码：48 / 59

页数：12

共 27 条

[1] NARX models of an industrial power plant gas turbine [J].