一种逆向样本分布的Boosting类新算法

化工学报 ›› 2011, Vol. 62 ›› Issue (8): 2287-2291.

一种逆向样本分布的Boosting类新算法

高敬阳，陈程立诏，朱群雄

北京化工大学信息科学与技术学院，北京 100029

出版日期:2011-08-05 发布日期:2011-08-05

A new Boosting algorithm：update training sample’s weight according to inverse error vector

GAO Jingyang，CHEN Chenglizhao，ZHU Qunxiong

Online:2011-08-05 Published:2011-08-05

摘要/Abstract

摘要：

对IB(Inverse Boosting)〖JP+1〗神经网络集成算法进行了研究，提出了IB算法的改进算法IB+算法。改进算法继承了IB算法的逆向样本分布调整策略，并在训练的过程中将部分已训练好的个体子网进行中间层网络集成，利用该中间层集成网络生成新的训练样本分布。实验结果表明，对于逆向权值分布的Boosting类算法，个体子网之间的关联度对网络集成后的泛化性能影响很小，减小个体网络的泛化误差将使集成后的泛化性能提高。

关键词: FONT-SIZE: 9pt, mso-ascii-font-family: Calibri, mso-ascii-theme-font: minor-latin, mso-fareast-theme-font: minor-fareast, mso-bidi-font-family: 宋体, mso-ansi-language: EN-US, mso-fareast-language: ZH-CN, mso-bidi-language: AR-SA">网络集成算法FONT-SIZE: 9pt, mso-ascii-theme-font: minor-latin, mso-fareast-theme-font: minor-fareast, mso-bidi-font-family: 宋体, mso-ansi-language: EN-US, mso-fareast-language: ZH-CN, mso-bidi-language: AR-SA, mso-fareast-font-family: 宋体, mso-hansi-font-family: 宋体" lang=EN-US>;FONT-SIZE: 9pt, mso-ascii-font-family: Calibri, mso-ascii-theme-font: minor-latin, mso-fareast-theme-font: minor-fareast, mso-bidi-font-family: 宋体, mso-ansi-language: EN-US, mso-fareast-language: ZH-CN, mso-bidi-language: AR-SA">逆向样本权值分布FONT-SIZE: 9pt, mso-ascii-theme-font: minor-latin, mso-fareast-theme-font: minor-fareast, mso-bidi-font-family: 宋体, mso-ansi-language: EN-US, mso-fareast-language: ZH-CN, mso-bidi-language: AR-SA, mso-fareast-font-family: 宋体, mso-hansi-font-family: 宋体" lang=EN-US>;FONT-SIZE: 9pt, mso-ascii-font-family: Calibri, mso-ascii-theme-font: minor-latin, mso-fareast-theme-font: minor-fareast, mso-bidi-font-family: 宋体, mso-ansi-language: EN-US, mso-fareast-language: ZH-CN, mso-bidi-language: AR-SA">中间层网络集成

Abstract:

This paper gives a research on IB(inverse Boosting)algorithm and proposes an improved version of IB called IB+.Both IB and IB+ algorithm will enhance the weight of samples which have been classified correctly during the training process.The most difference between IB and IB+ is the method to update the weight of training samples in each iteration.For IB algorithm,the weight of training samples will be updated according to an inverse error vector which was decided by the performance of the last trained single net.However the IB+ algorithm adopts a mesosphere ensemble net instead of a single net to determine the inverse error vector thus a more suitable sample distribution will be achieved.Further experiment results show that the performance of ensemble net which was developed using an inverse error vector to create new sample distribution will be decided by the performance of base single net not the degree of correlation.

Key words: FONT-SIZE: 9pt, mso-ascii-font-family: Calibri, mso-ascii-theme-font: minor-latin, mso-fareast-theme-font: minor-fareast, mso-bidi-font-family: 宋体, mso-ansi-language: EN-US, mso-fareast-language: ZH-CN, mso-bidi-language: AR-SA">网络集成算法FONT-SIZE: 9pt, mso-ascii-theme-font: minor-latin, mso-fareast-theme-font: minor-fareast, mso-bidi-font-family: 宋体, mso-ansi-language: EN-US, mso-fareast-language: ZH-CN, mso-bidi-language: AR-SA, mso-fareast-font-family: 宋体, mso-hansi-font-family: 宋体" lang=EN-US>;FONT-SIZE: 9pt, mso-ascii-font-family: Calibri, mso-ascii-theme-font: minor-latin, mso-fareast-theme-font: minor-fareast, mso-bidi-font-family: 宋体, mso-ansi-language: EN-US, mso-fareast-language: ZH-CN, mso-bidi-language: AR-SA">逆向样本权值分布FONT-SIZE: 9pt, mso-ascii-theme-font: minor-latin, mso-fareast-theme-font: minor-fareast, mso-bidi-font-family: 宋体, mso-ansi-language: EN-US, mso-fareast-language: ZH-CN, mso-bidi-language: AR-SA, mso-fareast-font-family: 宋体, mso-hansi-font-family: 宋体" lang=EN-US>;FONT-SIZE: 9pt, mso-ascii-font-family: Calibri, mso-ascii-theme-font: minor-latin, mso-fareast-theme-font: minor-fareast, mso-bidi-font-family: 宋体, mso-ansi-language: EN-US, mso-fareast-language: ZH-CN, mso-bidi-language: AR-SA">中间层网络集成

中图分类号:

TP 183

高敬阳, 陈程立诏, 朱群雄. 一种逆向样本分布的Boosting类新算法 [J]. 化工学报, 2011, 62(8): 2287-2291.

GAO Jingyang, CHEN Chenglizhao, ZHU Qunxiong. A new Boosting algorithm：update training sample’s weight according to inverse error vector[J]. CIESC Journal, 2011, 62(8): 2287-2291.

[1]	张艾萍, 谢立强, 徐志明. 超声空化对换热器协同场的影响 [J]. 化工学报, 2011, 62(S1): 130-133.
[2]	曾庆, 郭印诚, 牛振祺, 林文漪. 填料塔中氨水吸收二氧化碳的传质性能 [J]. 化工学报, 2011, 62(S1): 146-150.
[3]	褚燕彬, 曹柳林, 王晶. 利用组合B样条神经网络实现对间歇反应过程的动态建模 [J]. 化工学报, 2011, 62(S1): 157-162.
[4]	吴海波, 方岩雄, 纪红兵. 助溶剂促进的β-环糊精与肉桂醛的包结性能 [J]. 化工学报, 2011, 62(S1): 168-173.
[5]	王建立, 朱建军, 宋辰兴, 张兴. 3ω法测量纳米流体热导率 [J]. 化工学报, 2011, 62(S1): 42-47.
[6]	马维刚, 王海东, 张兴, 王玮. 多晶金薄膜电子-声子耦合的实验研究 [J]. 化工学报, 2011, 62(S1): 48-53.
[7]	崔福庆, 何雅玲, 程泽东, 李东, 陶于兵. 有压腔式吸热器内辐射传播过程的Monte Carlo模拟 [J]. 化工学报, 2011, 62(S1): 60-65.
[8]	何奎, 汪双凤, 黄间珍. 气液两相流在微小T型三通中的相分配特性 [J]. 化工学报, 2011, 62(S1): 92-96.
[9]	刘绩伟, 吴玉庭, 马重芳. 管肋式空间辐射器设计 [J]. 化工学报, 2011, 62(S1): 103-107.
[10]	李旺, 王情愿, 李瑞龙, 李超, 宇波, 张劲军, 代鹏飞. 大型浮顶油罐温度场数值模拟 [J]. 化工学报, 2011, 62(S1): 108-112.
[11]	胡朝发, 贾力. 脉动热管气液塞振荡运动模型 [J]. 化工学报, 2011, 62(S1): 113-117.
[12]	任楠, 王涛, 吴玉庭, 马重芳. 混合碳酸盐的DSC测量与比热容分析 [J]. 化工学报, 2011, 62(S1): 197-202.
[13]	马有福，袁益超，陈昱，刘海磊，康明. 翅片螺距对锯齿螺旋翅片换热管特性的影响 [J]. CIESC Journal, 2011, 62(9): 2484-2489.
[14]	刘戈，解茂昭，贾明. 基于互动小火焰模型的内燃机燃烧过程大涡模拟 [J]. CIESC Journal, 2011, 62(9): 2490-2498.
[15]	刘娇，孟范平，王震宇，刘启元. 亚甲基蓝光度法研究基于CaO₂的Fenton反应条件 [J]. CIESC Journal, 2011, 62(9): 2520-2526.