Low-rank approximations

Give matrix and a positive integer , we wish to find an matrix of rank at most ，so as to minimize the Frobenius norm of the matrix difference ，defined to be

Thus, the Frobenius norm of measures the discrepancy between and ; our goal is to find a matrix that minimizes this discrepancy, while constraining to have rank at most . If is the rank of , clearly and the Frobenius norm of the discrepancy is zero in this case. When is far smaller than , we refer to as a low-rank approximation.

SVD

  1. Given , constuct its SVD int the form of .
  2. Derive from the matrix formed by replacing by zeros the smallest sigular values on the diagnal of .
  3. Compute and output as the rank- approximation to .

http://nlp.stanford.edu/IR-book/html/htmledition/low-rank-approximations-1.html

Low-rank approximations的更多相关文章

  1. Generalized Low Rank Approximation of Matrices

    Generalized Low Rank Approximations of Matrices JIEPING YE*jieping@cs.umn.edu Department of Computer ...

  2. 吴恩达机器学习笔记59-向量化：低秩矩阵分解与均值归一化（Vectorization: Low Rank Matrix Factorization & Mean Normalization）

    一.向量化:低秩矩阵分解 之前我们介绍了协同过滤算法,本节介绍该算法的向量化实现,以及说说有关该算法可以做的其他事情. 举例:1.当给出一件产品时,你能否找到与之相关的其它产品.2.一位用户最近看上一 ...

  3. Sparse Principal Component Analysis via Regularized Low Rank Matrix Approximation(Adjusted Variance)

    目录 前言 文章概述 固定\(\widetilde{\mathrm{v}}\) 固定\(\widetilde{\mathrm{u}}\) Adjusted Variance 前言 这篇文章用的也是交替 ...

  4. 推荐系统(recommender systems):预测电影评分--构造推荐系统的一种方法：低秩矩阵分解(low rank matrix factorization)

    如上图中的predicted ratings矩阵可以分解成X与ΘT的乘积,这个叫做低秩矩阵分解. 我们先学习出product的特征参数向量,在实际应用中这些学习出来的参数向量可能比较难以理解,也很难可 ...

  5. 矩阵分解（rank decomposition）文章代码汇总

    矩阵分解(rank decomposition)文章代码汇总 矩阵分解(rank decomposition) 本文收集了现有矩阵分解的几乎所有算法和应用,原文链接:https://sites.goo ...

  6. <<Numerical Analysis>>笔记

    2ed,  by Timothy Sauer DEFINITION 1.3A solution is correct within p decimal places if the error is l ...

  7. [综述]Deep Compression/Acceleration深度压缩/加速/量化

    Survey Recent Advances in Efficient Computation of Deep Convolutional Neural Networks, [arxiv '18] A ...

  8. <Numerical Analysis>(by Timothy Sauer) Notes

    2ed,  by Timothy Sauer DEFINITION 1.3A solution is correct within p decimal places if the error is l ...

  9. 网络压缩论文集(network compression)

    Convolutional Neural Networks ImageNet Models Architecture Design Activation Functions Visualization ...

  10. CVPR 2015 papers

    CVPR2015 Papers震撼来袭! CVPR 2015的文章可以下载了,如果链接无法下载,可以在Google上通过搜索paper名字下载(友情提示:可以使用filetype:pdf命令). Go ...

随机推荐

  1. nodejs+express+mysql 增删改查(二)

    1.最早一篇关于express框架简单的增删改查文章,http://www.cnblogs.com/zhengyeye/p/nodejs.html#3947308:意外走红博客园,无奈自己之前一直没有 ...

  2. mongol学习

      启动mongodb 服务器:     进入mongodb文件夹:cd ~/mongodb     第一次先要创建set与log文件夹.   mkdir set; mkdir log;    并创建 ...

  3. java之设计模式工厂三兄弟之抽象工厂模式

    [学习难度:★★★★☆,使用频率:★★★★★]  工厂方法模式通过引入工厂等级结构,解决了简单工厂模式中工厂类职责太重的问题,但由于工厂方法模式中的每个工厂只生产一类产品,可能会导致系统中存在大量的工 ...

  4. [LeetCode] Open the Lock 开锁

    You have a lock in front of you with 4 circular wheels. Each wheel has 10 slots: '0', '1', '2', '3', ...

  5. [LeetCode] Minimum Index Sum of Two Lists 两个表单的最小坐标和

    Suppose Andy and Doris want to choose a restaurant for dinner, and they both have a list of favorite ...

  6. JS 变量类型互相转换

    转载自:http://zhaizhiyuan.blog.163.com/blog/static/1897672632009093147268/ Java中几种常用的数据类型之间转换方法:  1. sh ...

  7. UVA11082：Matrix Decompressing

    题意:给定一个矩阵的前i行的和,以及前i列的和,求任意一个满足条件的矩阵,矩阵元素在[1,20] 矩阵行列<=20 题解:做一个二分图的模型,把行列拆开,然后设源点到行节点的容量就是该行所有元素 ...

  8. 洛谷mNOIP模拟赛Day1-分组

    传送门 首先是贪心的思路 从后向前选,能多选就多选, 理由:数字越少肯定越优,同时间隔尽量向前推,字典序尽量小 对于K==1,枚举1~512直接判断 对于K==2,需要用镜像并查集,来刻画" ...

  9. bzoj1233[Usaco2009Open]干草堆tower 单调队列优化dp

    1233: [Usaco2009Open]干草堆tower Time Limit: 10 Sec  Memory Limit: 162 MBSubmit: 983  Solved: 464[Submi ...

  10. Miller-Rabin,Pollard-Rho(BZOJ3667)

    裸题直接做就好了. #include <cstdio> #include <cstdlib> #include <algorithm> using namespac ...