2021年12月20日上午9点,数学与统计学院在教学科研楼507室邀请博士研究生刘新蕊同学做线下学术报告,文体部负责人毛祥主持本次报告会,研究生一年级全员参加。
论文题目:Weighted Least Squares Support Vector Regression for complex censored regression
摘要:In order to overcome the di_culties of the regression estimations when the responses are subject to interval censoring or left truncation and right censoring, a least squares support vector regression(LS-SVR) is proposed based on multiple imputation and weighting. Single imputation and multiple imputation are carried out for _nite-interval censored (but not right censored) data respectively, the regression model was estimated by combining least squares support vector regression for right-censored data. For left-truncated and rightcensored data, a weight is proposed to reduce the e_ects of truncation and censoring on the least squares support vector regression procedure. Simulation results show that the proposed methods can e_ectively reduce the regression error and yield high accuracy than other least squares support vector regression does.
会议过程中,刘新蕊同学以其清晰的思路和严谨的思维行云流水地向硕士研究生们展示了其论文的主要思想和研究过程,即为了克服回归估计时的困难响应服从区间截尾或左截尾和右截尾,提出了最小二乘支持向量回归(LS-SVR)算法和一种加权算法关于最小二乘支持向量回归程序。通过统计方法研究,其论文模拟结果表明,所提出的方法能有效地减少回归误差。刘新蕊同学最终通过各个领域的实际数据验证了其研究的泛化能力,为今后类似的数理统计研究提供了很好的思路。
报告会结束后,在座同学针对基于多重输入和加权的最小二乘支持向量回归(LS-SVR)方法做出了深入思考,通过本次学术交流会,加强了我院的学术交流,让同学们,丰富学术知识,充分达到了这次报告会的目的,拓展了老师同学们的学术视野,激发了同学们的学习热情,同学们对用最小二乘回归角度处理删失数据的方式得到了启发,在场同学认识到了自身专业技能的不足并且体会到了亟待提升的实践能力。
数学与统计学院
2021年12月20日