构建复杂回响状态网络的新方法

宋青松; 冯祖仁

您当前的位置：

首页 >

文章列表页 >

构建复杂回响状态网络的新方法

更新时间：2025-07-09

- 构建复杂回响状态网络的新方法
- A New Method to Construct Complex Echo State Networks
- 西安交通大学学报 2009年43卷第4期页码：1-4
- 作者机构：
  
  西安交通大学系统工程研究所,西安,710049
- 作者简介：
- 基金信息：
  
  : 2008-07-08. 作者简介: 宋青松(1980-), 男,博士生;冯祖仁(联系人),男,教授,博士生导师. 基金项目: 国家重点基础研究发展规划资助项目(2007CB311006);国家高技术研究发展计划资助项目(2006AA04Z222)
- DOI：
  中图分类号： TP183
- 网络首发：2009-04-10，
  
  纸质出版：2009
- 稿件说明：
移动端阅览
宋青松, 冯祖仁. 构建复杂回响状态网络的新方法[J]. 西安交通大学学报, 2009,43(4):1-4.

A New Method to Construct Complex Echo State Networks[J]. 2009, 43(4): 1-4.
宋青松, 冯祖仁. 构建复杂回响状态网络的新方法[J]. 西安交通大学学报, 2009,43(4):1-4. DOI：

A New Method to Construct Complex Echo State Networks[J]. 2009, 43(4): 1-4. DOI：

摘要

研究了人工神经网络混沌时间序列预测应用中的拓扑选择问题

受大脑皮层生长发育过程启发

提出一种类皮层网络——复杂回响状态网络(CESN)的构建方法.将影响皮层网络生长发育过程的依赖距离和依赖时间窗口的2个生长机制中的规模因子、距离敏感因子、密度敏感因子、种子神经元个数

以及时间窗宽度等调控因子

类比地定义为CESN构建过程中的构造参数.实验发现

CESN的拓扑结构能够被这些构造参数惟一地确定

并且当种子神经元个数取值约为网络规模的10%时

构建出的CESN几乎具有最优的预测精度.根据这个发现

仿真证实了CESN能够被更省时地构建

表明所提方法是一种高效的网络构建方法.

Abstract

The issue of topology choosing in artificial neural networks for chaotic time series forecasting is studied. Motivated by the growth and development process of cerebral cortex

an algorithm for constructing cortex-like neural network

complex echo state network(CESN)

is proposed. Some factors in both the distance-dependent growth mechanism and time-window-dependent growth mechanism

which have effects on the process of the growth and development of the cortex networks

are analogously defined as construction parameters that controls the process in creating a CESN

such as the size factor

the distance-sensitive factor

the density-sensitive factor

the seed-neuron number and the time-window width

etc. It is found from experiments that the topological characters of a CESN can be uniquely determined by these parameters

and that when the seed-neuron number is set to be about 10% of the whole network size

the created CESN can have almost the best forecasting performance. Simulation results show that CESNs can be created with less time-consumption and better forecasting performance. It can be concluded that the proposed algorithm is a more efficient network constructing method.

关键词

Keywords

references

JAEGER H, HAASS H. Harnessing nonlinearity: predicting chaotic systems and saving energy in wireless communication [J]. Science, 2004, 304(5667):78-80.

王军平,陈全世,田光宇. 随机模糊神经网络及在随机混沌时间序列预测中的应用 [J]. 西安交通大学学报, 2003, 37(10):991-994.

WANG Junping, CHEN Quanshi, TIAN Guangyu. Stochastic fuzzy neural network and its application to prediction of stochastic chaotic time series [J]. Journal of Xi'an Jiaotong University, 2003, 37(10):991-994.

OZTURK M C, XU D, PRINCIPE J C. Analysis and design of echo state networks [J]. Neural Computation, 2007, 19(1):111-138.

EDELMAN G M, TONONI G. A universe of consciousness: how matter becomes imagination [M]. New York,USA: Basic Books, 2000.

KAISER M, HILGETAG C C. Development of multi-cluster cortical networks by time windows for spatial growth [J]. Neurocomputing, 2007, 70(10/12):1829-1832.

NISBACH F, KAISER M. Developmental time windows for spatial growth generate multiple-cluster small-world networks [J]. The European Physical Journal:B, 2007, 58(2): 185-191.

裴晓梅,郑崇勋,宾光宇. 基于多通道脑电特征运动意识任务的分类 [J]. 西安交通大学学报, 2005, 39(8):904-907.

PEI Xiaomei, ZHENG Chongxun, BIN Guangyu. Classification of the mental tasks based on multichannel electroencephalogram features [J]. Journal of Xi'an Jiaotong University, 2005, 39(8):904-907.

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

人体步态复杂度的递归图和递归定量分析研究

张角凹坑气膜孔对双层壁气膜冷却特性的影响

电动车辆间接热系统快速除湿响应迟滞机理及优化控制策略

冷库水平空气幕热湿阻隔性能数值模拟研究

离子液体改性燃料的电流体动力学效应研究