[1]何春梅,叶征春,韩牟,等.形态双向联想记忆网络的摄动鲁棒性研究[J].南京理工大学学报(自然科学版),2011,(05):664-669.
 HE Chun-mei,YE Zheng-chun,HAN Mu,et al.Perturbation Robustness of Morphological Bidirectional Associative Memories Networks[J].Journal of Nanjing University of Science and Technology,2011,(05):664-669.
点击复制

形态双向联想记忆网络的摄动鲁棒性研究
分享到:

《南京理工大学学报》(自然科学版)[ISSN:1005-9830/CN:32-1397/N]

卷:
期数:
2011年05期
页码:
664-669
栏目:
出版日期:
2011-10-31

文章信息/Info

Title:
Perturbation Robustness of Morphological Bidirectional Associative Memories Networks
作者:
何春梅1叶征春2韩牟3叶有培3
华东交通大学1. 信息工程学院; 2. 机电工程学院,江西南昌330013; 3. 南京理工大学计算机科学与技术学院,江苏南京210094
Author(s):
HE Chun-mei1YE Zheng-chun2HAN Mu3YE You-pei3
1. School of Information Engineering; 2. School of Mechanical and Electronical Engineering, East China Jiaotong University,Nanchang 330013,China; 3. School of Computer Science and Technology,NUST,Nanjing 210094,China
关键词:
神经网络 形态双向联想记忆 模糊 摄动 训练模式对 鲁棒性
Keywords:
neural networks morphological bidirectional associative memories fuzziness perturbation training pattern pairs robustness
分类号:
TP18
摘要:
针对训练模式对的小幅摄动可能会对神经网络的输出产生副作用的问题,该文研究了 形态双向联想记忆( MBAM) 网络和模糊形态双向联想记忆( FMBAM) 网络训练模式对的摄动鲁 棒性。比较了训练模式对摄动前后2 种网络输出范围的变化,理论证明了MBAM 和FMBAM 2 种网络的摄动鲁棒性。理论和仿真实例表明: MBAM 网络对训练模式对的摄动全局拥有好的鲁 棒性,其训练模式对的选取可以适度粗糙; 而FMBAM 网络对训练模式对的摄动不具有好的鲁 棒性,则其训练模式对的选取精度要求较高。
Abstract:
In view of the problem that the small perturbations of training pattern pairs may cause side effects on the outputs of neural networks, the influences of perturbations of training pattern pairs on the morphological bidirectional associative memories( MBAM) networks and fuzzy morphological bidirectional associative memories( FMBAM) networks are researched here. By comparing the output variations of the two training sample pairs before and after perturbations, the perturbation robustness of the MBAM networks and FMBAM networks is theoretically proved. The simulations show that the MBAM network has good robustness on the perturbation of training pattern pairs, and its rough training pattern pairs are acceptable. The FMBAM network robustness is not so good, and its training pattern pairs should have higher accuracy.

参考文献/References:

[1] Sussner P,Valle M E. Classification of fuzzy mathematical morphologies based on concepts of inclusion measure and duality[J]. Journal of Mathematical Imaging and Vision, 2008, 32( 2) : 139-159.
[2] Ritter G X,Sussner P,Diaz-de-Leon J L. Morphological associative memories[J]. IEEE Transactions on Neural Networks, 1998,9 ( 2) : 281-293.
[3] Ritter G X,Diaz-de-Leon J L,Sussner P. Morphological bidirectional associative memories [ J ]. Neural Network, 1999, 12( 6) : 851-867.
[4] 王敏,王士同,吴小俊. 新模糊形态学联想记忆网络 的初步研究[J]. 电子学报, 2003, 31( 5) : 690-693.
[5] 吴锡生,王士同. 模糊形态双向联想记忆网络的研 究[J]. 计算机工程, 2005, 31( 20) : 22-24.
[6] Valle M E,Sussner P. A general framework for fuzzy morphological associative memories[J]. Fuzzy Sets and Systems, 2008, 159( 7) : 747-768.
[7] 冯乃勤,刘春红,张聪品,等. 形态学联想记忆框架 研究[J]. 计算机学报, 2010, 33( 1) : 157-166.
[8] Zeng Xiaoqin,Wang Yingfeng,Zhang Kang. Computation of Adalines’Sensitivity to Weight Perturbation[J]. IEEE Transactions on Neural Networks, 2006, 17( 2) : 515-519.
[9] 徐蔚鸿,宋鸾姣,李爱华,等. 训练模式对的摄动对 模糊双向联想记忆网络的影响和控制[J]. 计算机 学报, 2006, 29( 2) : 337-344.
[10] He Chunmei,Ye Youpei,Xu Weihong. Research on perturbation of training pattern pairs on monolithic fuzzy neural networks[J]. Journal of Information Computational Science, 2007,4 ( 3) : 905-911.
[11] 何春梅,叶有培,徐蔚鸿. 训练模式对的摄动对单体 模糊神经网络的影响[J]. 南京理工大学学报, 2009, 33( 1) : 12-15.
[12] 曾水玲,徐蔚鸿,杨静宇. 训练模式摄动对模糊形态 学联想记忆网络的影响[J]. 模式识别与人工智能, 2010, 23( 1) : 91-96.

相似文献/References:

[1]刘永建,朱剑英,曾捷,等.改进BP神经网络在发动机性能趋势分析和故障诊断中的应用[J].南京理工大学学报(自然科学版),2010,(01):24.
 LIU Yong-jian,ZHU Jian-ying,ZENG Jie.Improved BP Neural Network System for Engine Performance Trend Analysis and Fault Diagnosis[J].Journal of Nanjing University of Science and Technology,2010,(05):24.
[2]黄俊,徐越兰.碳钢焊条熔敷金属力学性能非线性神经网络组合预测[J].南京理工大学学报(自然科学版),2012,36(05):800.
 HUANG Jun,XU Yue-lan.Nonlinear Combination Prediction of Mechanical Properties of CarbonSteel Electrode Deposited Metal Based on Neural Network[J].Journal of Nanjing University of Science and Technology,2012,36(05):800.
[3]袁军堂,夏玲玲,汪振华,等.灌胶结合面对数控机床动态性能影响研究[J].南京理工大学学报(自然科学版),2012,36(06):0.
 YUAN Jun tang,XIA Ling ling,WANG Zhen hua,et al.Impact of Glued Joint on Dynamic Characteristics of CNC Machine Tool[J].Journal of Nanjing University of Science and Technology,2012,36(05):0.
[4]朱 红,陈清华,刘国岁.高速神经网络HS-K-WTA-2的研究[J].南京理工大学学报(自然科学版),2007,(01):89.
 ZHU Hong,CHEN Qing-hua,LIU Guo-sui.High-speed Neural Network HS-K-WTA-2[J].Journal of Nanjing University of Science and Technology,2007,(05):89.
[5]都洪基,苏炜宏,邓烽.一种基于神经网络的谐波电流抑制方法[J].南京理工大学学报(自然科学版),2003,(01):28.
 DuHongji SuWeihong DengFeng.Adaptive Approach to Detection of the Harmonic Currents on Artificial Neural Network[J].Journal of Nanjing University of Science and Technology,2003,(05):28.
[6]魏荣,卢俊国,李军,等.α阶时延逆系统的小波神经网络实现[J].南京理工大学学报(自然科学版),2001,(04):342.
 WeiRong LuJunguo LiJun WangZhiquan.Realization of α-order Time-delay Inverse Systems Based on Wavelet Neural Networks[J].Journal of Nanjing University of Science and Technology,2001,(05):342.
[7]聂伟荣,朱继南,赵玉霞.基于改进BP网络的地震动信号目标识别[J].南京理工大学学报(自然科学版),2000,(01):20.
 NieWeirong ZhuJinan ZhaoYuxia.Microseismic Signal Targets Identification Based on Improved BP Neural Networks[J].Journal of Nanjing University of Science and Technology,2000,(05):20.
[8]卢云许,陆宝春,张世琪.切片平均分子量神经网络预测模型研究及应用[J].南京理工大学学报(自然科学版),2000,(02):139.
 LuYunxu LuBaochun ZhangShiqi.The Research and Application of Neural Network Predicting Model of Molecule Weight of Casting Slice Belt[J].Journal of Nanjing University of Science and Technology,2000,(05):139.
[9]牛玉刚,杨成梧,赵建丛.基于多步预测的PID型神经网络控制[J].南京理工大学学报(自然科学版),2000,(03):199.
 NiuYugang YangChengwu ZhaoJiancong.A PID-like Neural Network Control Based on Multi-step Prediction[J].Journal of Nanjing University of Science and Technology,2000,(05):199.
[10]赵启林,卓家寿,柳景春.BP网络的BP-G-S学习算法[J].南京理工大学学报(自然科学版),1999,(04):308.
 ZhaoQilin ZhuoJiashou LiuJingchuen.The BP G S Learning Algorithms of BP Network[J].Journal of Nanjing University of Science and Technology,1999,(05):308.

备注/Memo

备注/Memo:
基金项目: 华东交通大学科研基础项目( 09102018) ; 南京理工大学科研计划项目( ZYTS067) 作者简介: 何春梅( 1981-) ,女,博士,讲师,主要研究方向: 神经网络、模式识别理论及应用等,E-mail: xiaoxiao_he8 @163. com。
更新日期/Last Update: 2012-10-24