计算机工程 ›› 2012, Vol. 38 ›› Issue (10): 185-187.doi: 10.3969/j.issn.1000-3428.2012.10.056

• 人工智能及识别技术 • 上一篇    下一篇

改进BP神经网络在软件能耗分析中的应用

邬小龙 1,郭 兵 1,沈 艳 2   

  1. (1. 四川大学计算机学院,成都 610064;2. 成都信息工程学院控制工程学院,成都 610225)
  • 收稿日期:2011-06-01 出版日期:2012-05-20 发布日期:2012-05-20
  • 作者简介:邬小龙(1987-),男,硕士研究生,主研方向:嵌入式系统;郭 兵,教授、博士;沈 艳,副教授、博士
  • 基金项目:
    国家自然科学基金资助项目(61073045);四川省杰出青年科技基金资助项目(2010JQ0011)

Application of Improved BP Neural Network in Software Energy Consumption Analysis

WU Xiao-long 1, GUO Bing 1, SHEN Yan 2   

  1. (1. College of Computer Science, Sichuan University, Chengdu 610064, China;2. School of Control Engineering, Chengdu University of Information Technology, Chengdu 610225, China)
  • Received:2011-06-01 Online:2012-05-20 Published:2012-05-20

摘要: 提出以软件特征量为基础的嵌入式软件体系结构级能耗建模方法。利用软件特征量与嵌入式软件能耗之间存在非线性函数关系的特点,采用基于改进遗传算法的BP神经网络算法进行训练拟合。从初始群体、编码、模拟退火算子3个方面对遗传算法进行改进以增强拟合的效果。实验结果证明,基于该算法的能耗模型预测值与实际能耗值的误差较小。

关键词: 能耗建模, 软件特征量, 软件体系结构, 遗传算法, BP神经网络, 模拟退火算法

Abstract: This paper proposes a modeling of energy consumption in embedded software architecture level based on the software characteristic. It contains the idea that there is non-linear relationship between software characteristic and software energy consumption. And it presents a new BP neural network algorithm based on improved genetic algorithm to train and simulate the function. This algorithm takes three aspects into consideration to improve the genetic algorithm in order to enhance the simulation effect. Experimental results show that the deviation between the predictive value of energy consumption model based on this algorithm and the actual energy consumption value is smaller.

Key words: energy consumption modeling, software characteristic, software architecture, genetic algorithm, BP neural network, simulated annealing algorithm

中图分类号: