Distance between subspacess
2015-12-17 06:45
411 查看
Distance between subspaces
Definition of distance between subspaces
Theory of the distance
Conclusion of distance
Extension
Definition: Let E and F be d-dimensional subspaces of Rp with orthogonal projectors E and F. Denote the singular values of EF⊥ by <
4000
/span>s1,s2,⋯. The canonical angles between E and F are the numbers θk(E,F)=arcsin(sk)
for k=1,2,⋯,d and the angleoperator between E and F is the d×d matrix
Θ(E,F)=diag(θ1,θ2,⋯,θd).
∥sinΘ(E,F)∥2F=∑k=1ds2k=∥EF⊥∥2F=12∥E−F∥2F
If we also assume that VE and VF are the orthogonal bases for subspaces E and F respectively. Then from the Proposition 2.2 in “Minimax sparse principal subspace estimation in high dimension”, we have
12infQ∈Vd.d∥VE−VFQ∥2F≤∥sinΘ(E,F)∥2F≤infQ∈Vd.d∥VE−VFQ∥2F
This is to say the distance between two subspaces is equivalent to the minimal distance between their orthogonal bases.
∥VE−VF∥2F≤∥VEVtE−VFVtFProjection matrices∥2F
∥sinΘ(E,F)∥2F=12∥VEVtE−VFVtF∥2F≤⟨ΣE,VEVtE−VFVtF⟩λd(ΣE)−λd+1(ΣE)
∥sinΘ(E,F)∥2F=12∥VEVtE−VFVtF∥2F≤⟨ΣF,VFVtF−VEVtE⟩λd(ΣF)−λd+1(ΣF)
Combining these together,
∥VEVtE−VFVtF∥2F≤1δ⟨ΣE−ΣF,VEVtE−VFVtF⟩≤1δ∥ΣE−ΣF∥F∥VEVtE−VFVtF∥F
where δ=min{λd(ΣE)−λd+1(ΣE),λd(ΣF)−λd+1(ΣF)}
from which , we know
∥VEVtE−VFVtF∥F≤1δ∥ΣE−ΣF∥F.
Thus
∥VE−VFOrthogonal bases∥2F≤∥VEVtE−VFVtFProjection matrices∥2F≤1δ∥ΣE−ΣFSymmetric matrices∥F
Definition of distance between subspaces
Theory of the distance
Conclusion of distance
Extension
Distance between subspaces
1. Definition of distance between subspaces
How to describe the distance between subspaces? The canonical angles between subspaces can be used to define the distance between subspaces. And there are several equivalent ways to do this. Here I will talk is the definition used in the reference “Minimax sparse principal subspace estimation in high dimension”, which is based on the projection matrices of subpaces, because of the convenience to represent subspaces by there projector matrices.Definition: Let E and F be d-dimensional subspaces of Rp with orthogonal projectors E and F. Denote the singular values of EF⊥ by <
4000
/span>s1,s2,⋯. The canonical angles between E and F are the numbers θk(E,F)=arcsin(sk)
for k=1,2,⋯,d and the angleoperator between E and F is the d×d matrix
Θ(E,F)=diag(θ1,θ2,⋯,θd).
2. Theory of the distance
Since the nonzero singular values of E−F are the nonzero singular values of EF⊥, each counted twice. (which is the Proposition 2.1 in “Minimax sparse principal subspace estimation in high dimension”). Thus∥sinΘ(E,F)∥2F=∑k=1ds2k=∥EF⊥∥2F=12∥E−F∥2F
If we also assume that VE and VF are the orthogonal bases for subspaces E and F respectively. Then from the Proposition 2.2 in “Minimax sparse principal subspace estimation in high dimension”, we have
12infQ∈Vd.d∥VE−VFQ∥2F≤∥sinΘ(E,F)∥2F≤infQ∈Vd.d∥VE−VFQ∥2F
This is to say the distance between two subspaces is equivalent to the minimal distance between their orthogonal bases.
3. Conclusion of distance
Let VE∈Vp,d and EF∈Vp,d be two column orthogonal matrix, which is to say EtE=FtF=Id, with right rotation. Then the following is true,∥VE−VF∥2F≤∥VEVtE−VFVtFProjection matrices∥2F
4. Extension
If we assume that VE and VF span the d-dimensional principal subspaces (E,F) of ΣE and ΣF, then from the curvature lemma (lemma 4.2) given in “Minimax sparse principal subspace estimation in high dimension” , we have∥sinΘ(E,F)∥2F=12∥VEVtE−VFVtF∥2F≤⟨ΣE,VEVtE−VFVtF⟩λd(ΣE)−λd+1(ΣE)
∥sinΘ(E,F)∥2F=12∥VEVtE−VFVtF∥2F≤⟨ΣF,VFVtF−VEVtE⟩λd(ΣF)−λd+1(ΣF)
Combining these together,
∥VEVtE−VFVtF∥2F≤1δ⟨ΣE−ΣF,VEVtE−VFVtF⟩≤1δ∥ΣE−ΣF∥F∥VEVtE−VFVtF∥F
where δ=min{λd(ΣE)−λd+1(ΣE),λd(ΣF)−λd+1(ΣF)}
from which , we know
∥VEVtE−VFVtF∥F≤1δ∥ΣE−ΣF∥F.
Thus
∥VE−VFOrthogonal bases∥2F≤∥VEVtE−VFVtFProjection matrices∥2F≤1δ∥ΣE−ΣFSymmetric matrices∥F
相关文章推荐
- 10008---CSS Background背景
- 介绍几款Python IDE
- AngularJS Best Practices: resource
- 【纯感悟】我眼里的安卓架构师(一)
- 编译器对C函数的名字修饰
- ASA防火墙疑难杂症解答
- Project5&6 Verilog开发流水线CPU
- Is the type of “pointer-to-member-function” different from “pointer-to-function”?
- LeetCode "Maximum Product of Word Lengths"
- 关于大型网站技术演进的思考(二)--存储的瓶颈(2)
- C ++排序函数sort(),qsort()
- iis7设置dz论坛url静态化经验
- android自带标题样式
- 字典管理表空间 本地管理表空间 之间的区别(未懂)
- [LeetCode317]Shortest Distance from All Buildings
- [leetcode] 84. Largest Rectangle in Histogram 解题报告
- iGrimace V8 简单的安装方式
- python 线程及线程池
- 普法知识(40):宣传行政复议法的重要性
- 人品计算器