[1]王攀,李幼凤,冯 珊.模块化神经网络的Bayes子网集结新算法研究[J].智能系统学报,2006,1(2):79-83.
WANG Pan,LI You-feng,FENG Shan.Novel integrated algorithm of modular neural network’s?? sub-nets based on Bayesian learning[J].CAAI Transactions on Intelligent Systems,2006,1(2):79-83.
点击复制
WANG Pan,LI You-feng,FENG Shan.Novel integrated algorithm of modular neural network’s?? sub-nets based on Bayesian learning[J].CAAI Transactions on Intelligent Systems,2006,1(2):79-83.
模块化神经网络的Bayes子网集结新算法研究
《智能系统学报》[ISSN 1673-4785/CN 23-1538/TP] 卷:
1
期数:
2006年第2期
页码:
79-83
栏目:
学术论文—机器学习
出版日期:
2006-10-25
- Title:
- Novel integrated algorithm of modular neural network’s?? sub-nets based on Bayesian learning
- 文章编号:
- 1673-4785(2006)02-00079-05
- 分类号:
- TP183
- 文献标志码:
- A
- 摘要:
- 针对模块化神经网络的重要命题——子网动态集成问题，提出一种基于改进的Bayes学习的子网集结新方法.首先从处理复杂问题能力、计算开销、训练误差限等级的合理性、逼近正确率的构造等方面分析了已有方法的不足.既而提出相应策略，其核心在于采用了简洁、相关性小的子网生成方法；同时以误差作为依据提出新的逼近正确率指标以确定子网的动态集结权值.仿真实验对两种改进方法的测试误差进行了比较研究，结果表明了改进方法的有效性.
- Abstract:
- Aiming at the important issue of modular neural network (MNN) — the dynamic integration of the subnets, a novel integrated algorithm based on the improved Bayesian learning is presented. Firstly, the drawbacks of the old algorithm are analyzed from four aspectsthe processing ability for complex problems, compu ting cost, the rationality of trained error limit and the approximated accuracy. Then the corresponding strategy is presented whose key points are adopting a co ncise and lowly correlative subnets generating method, and then a new approxim a ted accuracy index based on the error measure is presented to determine subnet s ’ dynamic integration weights. The effectiveness of the above algorithms was de m onstrated by simulation through the comparative research to two improved algorit hms’ test accuracy.
- 参考文献/References:
-
［1】VALENTINI G MASULLI F. Ensembles of learning machines［A］. In: M. M arinaro and R. Tagliaferri (Eds.): WIRN VIETRI 2002［C］. LNCS 2486, 2002, 3-20.
［2］王 攀. 软计算方法在若干决策与控制问题中的应用研究［D］. 武汉: 华中科技大学, 2003.
?WANG Pan. Applicational reesearches with soft computing for some decision and c on trol issues［D］. Wuhan:Huazhong University of Science and Technology, 2 003.
［3］熊智华, 王 雄, 徐用懋. 基于分布式神经网络递推预报误差算法的非线性系统建模［J］. 信息与控制, 2000, 29 (5): 414-420.
XIONG Zhihua,WANG Xiong, XU Yongmao.Distributed neural networks based nonl inear system model using parallel recursive prediction error algorithm［J］. Inf ormat ion a nd Control， 2000, 29 (5): 414-420.
［4］蔡 骏. 面向决策支持的计算智能技术若干应用研究［D］. 武汉: 华中理工大学, 1996.
?CAI Jun. A study on the computational intelligence applied for decision supp ort ［D］. Wuhan:Huazhong University of Science and Technology,19 96
［5］AHMAD Z, ZHANG J. Bayesian selective combination [LL]of multiple neu ra l networ ks for improving longrange predictions in nonlinear process modeling［J］. Neu ral Comput & Applic，2005, 14: 78-87.
［6］李向武, 韦 岗. 基于小波网络的动态系统辨识方法及应用［J］. 控制理论与应用，1998， 15 (4):494-500.
LI Xiangwu,WEI gang. Dynamical system identification and its appl ications using wav elet network［J］. Chinese Control Theory and Applications，1998, 15(4): 494-500 .
备注/Memo
收稿日期：2006-01-04.
基金项目：国家自然科学基金资助项目(60174039)
作者简介：
王 攀，男，1971年生，毕业于华中理工大学，获工学博士学位. 现为武汉理工大学自动化学院副教授，控制与决策研究所所长.主要研究方向为智能优化与控制、决策分析，生物医学智能化系统.先后近10次获得省部市级科技进步、教研成果一、二、三等奖；发表论文50余篇.
李幼凤，女，1978年生，浙江大学信息学院博士研究生，主要研究方向为计算智能、预测控制.
冯 珊，女，1933年生，教授，博士生导师.清华大学本科，纽约州立大学石溪分校硕士.主要研究方向为智能决策支持系统，复杂系统建模与仿真.先后多次获得国家科技进步三等奖，省部级科技进步一、二等奖，发表论文100余篇.
更新日期/Last Update:
2009-05-05