Grassmann流形 — 用于子空间比较【理论】
1. 子空间
1.1 向量空间/线性空间
向量空间的定义
若一组向量组成的集合 V V V在数域 P P P上满足加法及乘法运算,则这组向量组成的集合 V V V被称为一个向量空间。
1.2 向量子空间/线性子空间
向量子空间定义
若集合 W W W是数域 P P P上向量空间 V V V的一个非空子集,且满足加法及乘法运算,则 W W W为 V V V的一个向量子空间。
2. Grassmann Manifolds
Mainfolds
Manifolds Described By Wikipedia
Definition
The Grassmannian manifold G ( m , D ) \mathcal{G}(m, D) G(m,D) refers to the D − d i m e n s i o n a l D-dimensional D−dimensional formed by all m − d i m e n s i o n a l m-dimensional m−dimensional embedded into a D − d i m e n s i o n a l D-dimensional D−dimensional real(or complex) Euclidean space.
An element of G ( m , D ) \mathcal{G}(m, D) G(m,D) can be represented by an orthonormal matrix Y Y Y of size D D D by m m m such that Y ′ Y = I m Y ^{'}Y = I_m Y′Y=Im, where I m I_m Im is the m m m by m m m identity matrix.
For example, Y Y Y can be the m m m basis vectors of a set of pictures in R D R^D RD.
However, the matrix representation of a point in G ( m , D ) \mathcal{G}(m, D) G(m,D) is not unique:
- Two matrices Y 1 Y_1 Y1 and Y 2 Y_2 Y2 are considered the same if and only if s p a n ( Y 1 ) = s p a n ( Y 2 ) span(Y_1) = span(Y_2) span(Y1)=span(Y2), where s p a n ( Y ) span(Y) span(Y) denotes the subspace spanned by the column vectors of Y Y Y.
- s p a n ( Y 1 ) = s p a n ( Y 2 ) span(Y_1) = span(Y_2) span(Y1)=span(Y2) if and only if Y 1 R 1 = Y 2 R 2 Y_1R_1 = Y_2R_2 Y1R1=Y2R2 for some R 1 , R 2 ∈ O ( m ) R_1, R_2 ∈ O(m) R1,R2∈O(m).
With this understanding, we will often use the notation Y Y Y when we actually mean its equivalence class s p a n ( Y ) span(Y) span(Y), and use Y 1 = Y 2 Y_1 = Y_2 Y1=Y2 when we mean s p a n ( Y 1 ) = s p a n ( Y 2 ) span(Y_1) = span(Y_2) span(Y1)=span(Y2