A NONMONOTONE TRUST REGION METHOD FOR NONLINEAR LEAST SQUARES PROBLEMS

In this paper we present a nonmonotone trust region method for nonlinear least squares problems with zero-residual and prove its convergence properties. The extensive numerical results are reported which show that the nonmonotone trust region method is generally superior to the usual trust region method.     

非对称相异性矩阵的一种非度量多维尺度变换

本文将Okada ＆ Imaizumi等的模型加以推广,提出了一种用于处理非对称相异性矩阵的非度量多维尺度变换新方法.在模型中,我们假定每个研究对象可以表示为Minkowski度量空间中的一个点和一个超球面,超球面的半径揭示了相应研究对象的非对称性.文中我们给出了一种计算点坐标及球半径的算法.该算法使用了代数方法,比原来的方法收敛速度快,节省计算时间.最后给出了一个数值例子.     

THE CONVERGENCE OF A DOMAIN DECOMPOSITION ALGORITHM FOR PARABOLIC PROBLEMS

At recent, Hourgat et gave a domain decomposition algorithm for elliptic problems which can be implemented in parallel. Many numerical experiments have illustrated its efficiency. In the present paper, we apply this algorithm to solve the discrete parabolic problems, analyse its convergence and show that its convergence rale is about (1 - 2p + σp2 ) which is nearly optimal and independent of the parameter τ, where σ τ O((1 +H )(1 + ln(H / h))2 ). 0 < p < 1 / σ,τ,h,H are the time step size, finite element parameter and subdomain diameter, respectively.     

多个相异性矩阵的数据分析

本文给出了分析多个相异性矩阵的三种方法.首先找到了一种图表示,使我们对所有相异性矩阵有一个总体的了解;其次定义了一个新的相异性矩阵,它可以看作是对所有原始相异性矩阵的一个折衷处理;最后提出了一种MIMU方法.在文中我们还对由上述方法得到的坐标图进行了比较.     

NONMONOTONE CURVED SEARCH METHODS FOR UNCONSTRAINED OPTIMIZATION

In this paper two nonmonolone curved search (NCS) algorithms fur unconstrained optimization are presented. The NCS algorithms possess both a global convergence properly and a quadratic rale of convergence. Some numerical results are also reported which show that the NCS algorithn is superior to the usual curved search (UCS)aIgorithm for typical lest problems.     

一种加权IDIOSCAL模型的非度量分析方法

§1 引言 IDIOSCAL模型是INDSCAL模型的推广,以前关于这方面的工作做得很少。IDIOSCAL(Individual Difference In Orientation SCALing)模型最早是由Carroll ＆ Chang(1970)提出来的,后来Tucker(1972)及Harshman(1972,1980)在这方面作了一些研究工作,除