摘要
作为机器学习法的一种,即时学习法(Just-in-time Learning, JITL)因其具有自适应特性和良好的预测能力,近年来受到国内外学者的广泛关注。本文简述了即时学习法的基本思想,对目前文献中已存在的即时学习法进行了总结,并综述了即时学习法在过程工业方面应用的最新研究进展。通过对即时学习法相关研究成果的总结,发现影响即时学习法性能的主要因素为选取不同近邻准则以及模型结构,目前主要的应用研究热点在于化工过程控制和软测量建模,最后指出今后可能的发展方向为即时学习法算法的改进和其应用领域的扩展。
As a method of machine learning, the just-in-time learning(JITL) method has been received widespread attentions due to its adaptive nature and good prediction capability in recent years. In this paper, the basic idea of JITL is described, the existing JITL methods in the literature are summarized, and the JITL applications in process industry are overviewed. Through the summary of relevant research results of JITL, it is concluded that the selections of different nearest neighborhood criterions and model structures are the main factors which affect the performance of JITL, and JITL is mainly used in process control and soft sensor. Finally,it is pointed that the improvement of JITL’s algorithm and the extension of application filed are the possible research developments of JITL.
作者
杨鑫
周成宇
YANG Xin;ZHOU Chengyu(School of Chemical Engineering,Chongqing University of Technology,Chongqing,400054,China)
出处
《计算机与应用化学》
CAS
北大核心
2018年第9期746-758,共13页
Computers and Applied Chemistry
基金
国家自然科学基金青年科学基金资助项目(21306234)
关键词
即时学习法
过程工业
过程控制
软测量
just-in-time learning
process industry
process control
soft sensor
