优化建模技术和机器学习理论的新发展  

New progress of technology of modelling and theory of machine learning

作　　者：陈念贻[1] 陆文聪 陆治荣 

机构地区：[1]上海大学理学院化学系计算机化学研究室,上海200436 [2]思华数据处理技术有限公司,宁波315040

出　　处：《计算机与应用化学》2002年第6期677-682,共6页Computers and Applied Chemistry

摘　　要：化工、炼油、冶金等制造业生产过程、新产品研制,以及经营管理的优化能给企业带来巨大的经济效益。优化成功的前提是需要建立能预报优化控制条件的数学模型。用机器学习技术从已有的数据中抽提出有用信息,是建立有效数学模型的关键。本文回顾了优化建模技术及其理论基础的几个发展阶段,指出从线性建模到非线性建模,从追求经验风险极小化到追求实际风险极小化,从采用单一算法到建立多种算法相结合的信息处理平台,从单纯根据古典统计数学到参照新发展的统计学习理论,使优化建模技术由粗到精,由低级到高级,在生产过程、新产品研制和经营管理的优化中发挥更大作用。The optimization of industnal production process, preparation of new products and enterprise management can make large economic profit for the enterprises dealing with the production of chemical, petrochemical or metallurgical products. The key problem of optimization works is to make models for the prediction of optimal condition of production, testing or management. So it is necessary to extract useful infor-mation from known data sets. In this paper, the history of development of optimization technology is reviewed. It is emphasized that the change from linear modeling to nonlinear modeling methods, the change from empirical risk minimization to real risk minimization, and the change from traditional statistical mathematics to newly proposed statistical learning theory, are the chief trends making modelling and optimization technolo-gy more advanced and more effective for improvement of the works in modern enterprises.

关 键 词：优化建模技术 机器学习理论 发展 过拟合 数学模型 模拟识别优化算法 支持向量机算法 智能控制 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP301.6[自动化与计算机技术—控制科学与工程]