|
|||||
|
|
| 一种基于脑网络的分层模块化回声状态网络 | |
| 引用本文： | 李丁园,刘富,乔俊飞. 一种基于脑网络的分层模块化回声状态网络[J]. 机械工程学报, 2015, 51(22): 128-134. DOI: 10.3901/JME.2015.22.128 |
| 作者姓名： | 李丁园 刘富 乔俊飞 |
| 作者单位： | 1. 吉林大学通信工程学院长春130022； 2. 北京工业大学电子信息与控制工程学院北京100124； 3. 北京工业大学计算智能与智能系统北京市重点实验室北京100124 |
| 基金项目： | 国家自然科学基金资助项目(61034008，61203099，61225016) |
| 摘 要： | 针对回声状态网络(Echo state network, ESN)结构设计问题，提出一种基于脑网络的分层模块化回声状态网络(Hierarchical modular echo state network, HMESN)。脑网络的拓扑结构使功能网络具有丰富的动力学特性，因此，从生物仿生学角度出发，对HMESN的储备池进行分层设计，各层级上的神经元采用小世界网络构建算法生成模块化结构，并引入层级连接。基于脑网络分层模块化的拓扑特征弱化了神经元间的耦合程度，从而使神经元的动力学特性更为丰富，在功能与结构上更接近于真实生物神经网络，有效地提高了网络处理问题的能力。采用Mackey-Glass时间序列预测和非线性系统辨识对网络进行验证，证明该网络的有效性和可行性。 |
| 关 键 词： | 储备池 回声状态网络 脑网络 时间序列预测 |
| 收稿时间： | 2015-02-28 |
A Hierarchical Modular Echo State Network Based on Brain Networks |
|
| LI Dingyuan,LIU Fu,QIAO Junfei. A Hierarchical Modular Echo State Network Based on Brain Networks[J]. Chinese Journal of Mechanical Engineering, 2015, 51(22): 128-134. DOI: 10.3901/JME.2015.22.128 | |
| Authors: | LI Dingyuan LIU Fu QIAO Junfei |
| Affiliation: | 1. College of Communication Engineering, Jilin University, Changchun 130022;2. College of Electronic Information and Control Engineering, Beijing University Technology, Beijing 100124;3. Beijing Key Laboratory of Computational Intelligence and Intelligent System, Beijing University Technology, Beijing 100124 |
| Abstract: | For the structure design of echo state network (ESN), a novel hierarchical modular echo state network (HMESN) based on brain networks is established. The topology of brain networks makes functional networks with rich dynamic behavior. Therefore, from a biological bionic point of view, the reservoir of HMESN for hierarchical design, each level of neurons using small world network construction algorithm to generate modular structure, as well as considered the hierarchical connection .Based on the topological characteristics of hierarchical modular in brain network, it weakens the coupling intensity among neurons, and enriches the dynamics of internal neurons. HMESN’s reservoir is closer to the real biological neural networks from both structural and functional perspectives, which effectively improves the information processing ability of network. Finally, this proposed network is used for the Mackey-Glass time series prediction and the non-linear system identification to prove its validity and feasibility. |
| Keywords: | brain networks echo state network reservoir time series prediction |
| 本文献已被 CNKI 万方数据 等数据库收录！ | |
| 点击此处可从《机械工程学报》浏览原始摘要信息 | |
| 点击此处可从《机械工程学报》下载全文 | |