基于优化ELM网络的物理量回归方法研究  

Research on physical quantity regression method based on optimized ELM network

作　　者：王平 王宜怀[2,4] 刘长勇 彭涛 WANG Ping;WANG Yihuai;LIU Changyong;PENG Tao(School of Mathematics and Computer Science,Wuyi University,Wuyishan 354300,China;School of Computer Science and Technology,Soochow University,Suzhou 215006,China;The Key Laboratory of Cognitive Computing and Intelligent Information Processing of Fujian Education Institutions,Wuyishan 354300,China;Collaborative Innovation Center of Novel Software Technology and Industrialization,Suzhou 215006,China)

机构地区：[1]武夷学院数学与计算机学院,福建武夷山354300 [2]苏州大学计算机科学与技术学院,江苏苏州215006 [3]认知计算与智能信息处理福建省高校重点实验室,福建武夷山354300 [4]软件新技术与产业化协同创新中心,江苏苏州215006

出　　处：《现代电子技术》2020年第17期141-146,共6页Modern Electronics Technique

基　　金：国家自然科学基金资助项目(61672369);中央引导地方科技发展专项资金项目(2018L3013);福建省教育厅科研基金资助项目(JA15522);武夷学院校科研基金资助项目(xl201016);福建省本科高校教育教学改革项目(FBJG20190281)。

摘　　要：针对传统的A/D值转换物理量回归方法中存在表达不统一、动态适应性弱和在线非线性校正能力不足等问题,尝试将机器学习的ELM网络引入到该应用中。在分析A/D值转换物理量回归的知识要素基础上,依托ELM网络的非线性映射能力,提出利用遗传算法优化ELM网络,并利用其实现统一数学表达的A/D值转换物理量回归方法。实际应用表明,该方法对物理量回归问题可实现统一的数学模型表达,泛化性好,且非线性校正能力强,实现了各类A/D值转换物理量回归应用。In view of the problems of traditional A/D value conversion physical quantity regression method,such as inconsistent expressions,poor dynamic adaptability and insufficient online nonlinear correction ability,the ELM(extreme learning machine)network of machine learning is introduced into this application. On the basis of the analysis on knowledge elements of A/D value conversion physical quantity regression,and relying on the nonlinear mapping ability of ELM network,an A/D value conversion physical quantity regression method is proposed,which uses genetic algorithm to optimize ELM network,so as to achieve consistent mathematical expressions. The practical application shows that the method can be used to achieve consistent mathematical model expressions for physical quantity regression and are of good generalization. In addition,it is of excellent nonlinear correction ability,so the regression application of various A/D value conversion physical quantities can be realized satisfactorily.

关 键 词：机器学习算法 模/数转换 极限学习机网络 遗传算法 优化方法 物理量回归 动态校正 

分 类 号：TN711-34[电子电信—电路与系统] TP391[自动化与计算机技术—计算机应用技术]