[1]许敏,史荧中,葛洪伟,等.一种具有迁移学习能力的RBF-NN算法及其应用[J].智能系统学报,2018,13(06):959-966.[doi:10.11992/tis.201705021]
 XU Min,SHI Yingzhong,GE Hongwei,et al.A RBF-NN algorithm with transfer learning ability and its application[J].CAAI Transactions on Intelligent Systems,2018,13(06):959-966.[doi:10.11992/tis.201705021]
点击复制

一种具有迁移学习能力的RBF-NN算法及其应用(/HTML)
分享到:

《智能系统学报》[ISSN:1673-4785/CN:23-1538/TP]

卷:
第13卷
期数:
2018年06期
页码:
959-966
栏目:
出版日期:
2018-10-25

文章信息/Info

Title:
A RBF-NN algorithm with transfer learning ability and its application
作者:
许敏12 史荧中2 葛洪伟1 黄能耿2
1. 江南大学 物联网技术学院, 江苏 无锡 214122;
2. 无锡职业技术学院 物联网技术学院, 江苏 无锡 214121
Author(s):
XU Min12 SHI Yingzhong2 GE Hongwei1 HUANG Nenggeng2
1. School of Internet of things technology, Jiangnan University, Wuxi 214122, China;
2. School of Internet of things technology, Wuxi Institute of Technology, Wuxi 214121, China
关键词:
径向基函数神经网络迁移学习径向基函数中心向量ε不敏感损失函数信息缺失
Keywords:
radial basis function neural networktransfer learningradial basis function vectorε-insensitive loss functionmissing information
分类号:
TP391
DOI:
10.11992/tis.201705021
摘要:
经典的径向基人工神经网络学习能逼近任意函数,因而应用广泛。但其存在的一个重要缺陷是,在已标签样本过少、不能反映数据集整体分布情况下,容易产生过拟合现象,从而导致泛化性能严重下降。针对上述问题,探讨具有迁移学习能力的径向基人工神经网络学习算法,该算法在引入ε不敏感损失函数和结构风险项的同时,学习源领域径向基函数的中心向量及核宽和源领域模型参数,通过充分学习历史源领域知识来弥补当前领域因已标签样本少而导致泛化能力下降的不足。将该算法应用于人造数据集和真实发酵数据集进行验证,和传统的RBF神经网络算法相比,所提算法在已标签样本少而存在数据缺失的场景下,具有更好的适应性。
Abstract:
The classical radial basis function neural network (RBF-NN) is widely used as it can approximate any function. However, one of its main defects is that overfitting is likely to occur when there are too few labeled samples to reflect the overall distribution of datasets; this leads to a serious decline in its generalization ability. To solve the above problem, an artificial RBF-NN learning algorithm with transfer learning ability is discussed. The algorithm introduces the ε-insensitive loss function and the structural risk term and also learns the center vector and kernel width of the radial basis function as well as the parameters of the source domain model. The algorithm fully learns the knowledge in the historical source domain to compensate for its decline in generalization ability caused by the lack of labeled samples in the current field. To verify the algorithm, it is applied to an artificial dataset and real fermentation dataset. Compared with the traditional RBF-NN algorithm, the proposed algorithm has a better adaptability as regards less labeled samples and missing data.

参考文献/References:

[1] MOODY J, DARKEN C J. Fast learning in networks of locally-tuned processing units[J]. Neural computation, 1989, 1(2):281-294.
[2] RYU D, LIANG Faming, MALLICK B K. Sea surface temperature modeling using radial basis function networks with a dynamically weighted particle filter[J]. Journal of the American statistical association, 2013, 108(501):111-123.
[3] 李方伟, 郑波, 朱江, 等. 一种基于AC-RBF神经网络的网络安全态势预测方法[J]. 重庆邮电大学学报:自然科学版, 2014, 26(5):576-581 LI Fangwei, ZHENG Bo, ZHU Jiang, et al. A method of network security situation prediction based on AC-RBF neural network[J]. Journal of Chongqing university of posts and telecommunications:natural science edition, 2014, 26(5):576-581
[4] 樊劲辉, 贾松敏, 李秀智. 基于RBF神经网络的全向智能轮椅自适应控制[J]. 华中科技大学学报:自然科学版, 2014, 42(2):111-115 FAN Jinhui, JIA Songmin, LI Xiuzhi. Adaptive control for omni-directional intelligent wheelchairs based on RBF neural network[J]. Journal of Huazhong university of science and technology:nature science edition, 2014, 42(2):111-115
[5] STASINAKIS C, SERMPINIS G, THEOFILATOS K, et al. Forecasting us unemployment with radial basis neural networks, Kalman filters and support vector regressions[J]. Computational economics, 2016, 47(4):569-587.
[6] PRATHIBA R, BALASINGHMOSES M, DEVARAJ D, et al. Multiple output radial basis function neural network with reduced input features for on-line estimation of available transfer capability[J]. Control engineering and applied informatics, 2016, 18(1):95-106.
[7] ALI S H A, OZAWA S, NAKAZATO J, et al. An online malicious spam email detection system using resource allocating network with locality sensitive hashing[J]. Journal of intelligent learning systems and applications, 2015, 7(2):55866.
[8] 韩敏, 穆云峰. 一种改进的RAN网络结构优化算法[J]. 控制与决策, 2007, 22(10):1177-1180 HAN Min, MU Yunfeng. Improved learning algorithm for optimizing RAN network structure[J]. Control and decision, 2007, 22(10):1177-1180
[9] PLATT J. A resource-allocating network for function interpolation[J]. Neural computation, 1991, 3(2):213-225.
[10] SARIMVEIS H, DOGANIS P, ALEXANDRIDIS A. A classification technique based on radial basis function neural networks[J]. Advances in engineering software, 2006, 37(4):218-221.
[11] RAITOHARJU J, KIRANYAZ S, GABBOUJ M. Training radial basis function neural networks for classification via class-specific clustering[J]. IEEE transactions on neural networks and learning systems, 2016, 27(12):2458-2471.
[12] PEDRYCZ W. Conditional fuzzy clustering in the design of radial basis function neural networks[J]. IEEE transactions on neural networks, 1998, 9(4):601-612.
[13] LACERDA E, DE CARVALHO A, LUDERMIR T. Evolutionary optimization of RBF networks[J]. International journal of neural systems, 2001, 11(3):287-294.
[14] SHEKHAR S, AMIN M B. Generalization by neural networks[J]. IEEE transactions on knowledge and data engineering, 1992, 4(2):177-185.
[15] ALEXANDRIDIS A, CHONDRODIMA E, SARIMVEIS H. Radial basis function network training using a nonsymmetric partition of the input space and particle swarm optimization[J]. IEEE transactions on neural networks and learning systems, 2013, 24(2):219-230.
[16] PAN S J, YANG Qiang. A survey on transfer learning[J]. IEEE transactions on knowledge and data engineering, 2010, 22(10):1345-1359.
[17] 张雅俊, 高陈强, 李佩, 等. 基于卷积神经网络的人流量统计[J]. 重庆邮电大学学报:自然科学版, 2017, 29(2):265-271 ZHANG Yajun, GAO Chenqiang, LI Pei, et al. Pedestrian counting based on convolutional neural network[J]. Journal of Chongqing university of posts and telecommunications:natural science edition, 2017, 29(2):265-271
[18] 桑庆兵, 邓赵红, 王士同, 等. 基于ε-不敏感准则和结构风险的鲁棒径向基函数神经网络学习[J]. 电子与信息学报, 2012, 34(6):1414-1419 SANG Qingbing, DENG Zhaohong, WANG Shitong, et al. ε-insensitive criterion and structure risk based radius-basis-function neural-network modeling[J]. Journal of electronics & information technology, 2012, 34(6):1414-1419
[19] 邓乃扬, 田英杰. 支持向量机:理论、算法与拓展[M]. 北京:科学出版社, 2009:63-80.
[20] 蒋亦樟, 邓赵红, 王士同. ML型迁移学习模糊系统[J]. 自动化学报, 2012, 38(9):1393-1409 JIANG Yizhang, DENG Zhaohong, WANG Shitong. Mamdani-Larsen type transfer learning fuzzy system[J]. Acta automatica sinica, 2012, 38(9):1393-1409

相似文献/References:

[1]王晓初,包芳,王士同,等.基于最小最大概率机的迁移学习分类算法[J].智能系统学报,2016,11(1):84.[doi:10.11992/tis.201505024]
 WANG Xiaochu,BAO Fang,WANG Shitong,et al.Transfer learning classification algorithms based on minimax probability machine[J].CAAI Transactions on Intelligent Systems,2016,11(06):84.[doi:10.11992/tis.201505024]
[2]王跃,杨燕,王红军.一种基于少量标签的改进迁移模糊聚类[J].智能系统学报,2016,11(3):310.[doi:10.11992/tis.201603046]
 WANG Yue,YANG Yan,WANG Hongjun.An improved transfer fuzzy clustering with few labels[J].CAAI Transactions on Intelligent Systems,2016,11(06):310.[doi:10.11992/tis.201603046]
[3]张艳霞,陈丹琪,韩莹,等.基宽灵敏度分析的径向基神经网络代理模型[J].智能系统学报,2014,9(02):259.[doi:10.3969/j.issn.1673-4785.201309009]
 ZHANG Yanxia,CHEN Danqi,HAN Ying,et al.Surrogate model of radial basis function networks based on width factor sensitivity analysis[J].CAAI Transactions on Intelligent Systems,2014,9(06):259.[doi:10.3969/j.issn.1673-4785.201309009]
[4]程旸,蒋亦樟,钱鹏江,等.知识迁移的极大熵聚类算法及其在纹理图像分割中的应用[J].智能系统学报,2017,12(02):179.[doi:10.11992/tis.201603005]
 CHENG Yang,JIANG Yizhang,QIAN Pengjiang,et al.A maximum entropy clustering algorithm based on knowledge transfer and its application to texture image segmentation[J].CAAI Transactions on Intelligent Systems,2017,12(06):179.[doi:10.11992/tis.201603005]
[5]莫宏伟,汪海波.基于Faster R-CNN的人体行为检测研究[J].智能系统学报,2018,13(06):967.[doi:10.11992/tis.201801025]
 MO Hongwei,WANG Haibo.Research on human behavior detection based on Faster R-CNN[J].CAAI Transactions on Intelligent Systems,2018,13(06):967.[doi:10.11992/tis.201801025]
[6]凌彤,杨琬琪,杨明.利用多模态U形网络的CT图像前列腺分割[J].智能系统学报,2018,13(06):981.[doi:10.11992/tis.201806012]
 LING Tong,YANG Wanqi,YANG Ming.Prostate segmentation in CT images with multimodal U-net[J].CAAI Transactions on Intelligent Systems,2018,13(06):981.[doi:10.11992/tis.201806012]

备注/Memo

备注/Memo:
收稿日期:2017-05-17。
基金项目:国家自然科学基金项目(61572236);江苏省高等学校自然科学研究项目(18KJB520048);江苏高校“青蓝工程”项目(苏教师〔2016〕15号);江苏省“333高层次人才培养工程”项目(苏人才〔2016〕7号).
作者简介:许敏,女,1980年生,副教授,博士,主要研究方向为人工智能、模式识别,发表学术论文10余篇;史荧中,男,1970年生,副教授,博士,主要研究方向为人工智能、模式识别,参与多项省级以上科研项目,发表学术论文10余篇;葛洪伟,男,1967年生,教授,博士生导师,博士,主要研究方向为人工智能、模式识别、机器学习、图像处理与分析等。主持和承担国家自然科学基金等国家级项目和省部级项目近20项,获省部级科技进步奖多项。发表学术论文百余篇。
通讯作者:许敏.E-mail:applexu9027@126.com
更新日期/Last Update: 2018-12-25