基于黎曼流形上的优化方法求解稀疏PCA.pdfVIP

英 文 摘 要 A b s tra c t PC A is a popu lar tool in data analysis, dim ension reduction and visualization. Sparse P C A is an im proved m ethod in order to overcom e the difficulty of the in terpre? tation of principal com ponent in the PC A calcu lation . Sparse PC A is a m anifo ld con? strained optim ization problem . H ow ever, the effect is not very w ell w hen calculating the prob lem by taking the spherical constraint as an equality constraint. In this paper, w e attem pt to transform the non linear constrained op tim ization problem into an opti? m ization prob lem on R iem annian m anifolds. W e also establish a line-search m ethod on R iem ann ian m anifold s to solve it. D uring th e process of algorithm bu ilding, w e g ive th e d e term in a tio n an d ca lcu la tio n m eth o d o f th e d e sc e n t d ire c tio n o f n o n -sm o o th fu n c ? tion s on R iem ann ian m an ifo ld in detail. A t last, th is pap er verify the convergence and feasib ility of our algorithm by num erical experim ents. K ey W ords: Sparse PC A , R iem annian M an ifolds, L ine-search M ethod, D escent D i? re c tio n C h in ese L ib rary C lassification nu m b er : 0 24 1.6 目 录 ：摘 要  II A b stra c t  Ill 第一章 绪论  1 1.1 稀 疏 P C A 问题 背 景 介 绍  1 1.2 本 文 结 构  3 1.3 符 号说 明 3 第二章 黎 曼流形上广义梯度 的概 念 与相 关性质  4 2.1 黎 曼流形及 相 关符 号介 绍 