Scribd
Upload
9 views8 pages

Singular Value Decomposition

Uploaded by

Shubham Wagh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
9 views8 pages

Singular Value Decomposition

Uploaded by

Shubham Wagh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

Singular Value Decomposition :

30 March 2021 00:00

Singular values of Matrix :


Let A be any m x n matrix.
Then the matrix AtA (Gram matrix)is a symmetric matrix of order n x n and can be orthogonally
diagonalisable.
Let, ƛ2, ƛ3, …..ƛn be Eigen values of AtA and v1, v2,…..vn be corresponding orthonormal eigen
vectors .
Now;
||Avi||2 = <Avi, Avi>
= (Avi)t . (Avi)
= vit. (AtA). Vi
= Vit ( ƛi. Vi)
= ƛi . (Vit. Vi)
= ƛi . <Vi, Vi>
= ƛi
So Eigen values of AtA are non-negative.
The singular values of A are square roots of Eigen values of AtA denoted by σ1, σ2, σ3….σn.
σ1= ƛ1, σ2 = ƛ2, σ3 = ƛ3….σn = ƛn.
i.e σi= ƛi
In other words Singular values of A are lengths of vectors AV1, AV2,…..

Example:

Note : If A has r non zero singular values then rank(A) = r.

Singular Value Decomposition :


Let A be an m x n matrix with rank r.
Then there exist an m x n matrix
Σ= and
an m x m orthogonal matrix U and an n x n orthogonal matrix V such that A = UΣVt.

The columns of U are called as left singular vectors of A and the columns of V are called as right
singular vectors of A.

Singular Value Decomposition Page 1


singular vectors of A.

Steps in SVD:
1. Find an orthogonal diagonalization of AtA.
That is , find Eigen values of AtA and the corresponding orthonormal set of Eigen vectors.
Construct V = [v1 | v2 |……|Vn]….where V1, V2,…..vn are orthonormal Eigen vectors of AtA
2. Construct Σ =
3. Construct U.
The column vectors of U are normalised vectors AV1, AV2,…..AVn.
U1 = σ1 σ2

Singular Value Decomposition Page 2


Singular Value Decomposition Page 3
Singular Value Decomposition Page 4
Singular Value Decomposition Page 5
Singular Value Decomposition Page 6
Reduced SVD and Pseudo inverse:
Let A be a matrix of order m x n and rank(A) = r.
Let A = UΣVt be a SVD of A.
Consider Ur = [u1 | u2 |……| ur]
Vr =[v1 | v2 |……|vr]
And D be a diagonal matrix whose diagonal elements are nonzero
singular values of A.
Then A can be written as ,
A = Ur D Vrt
Where Ur is a matrix of order m x r , D is a matrix of order r x r
and Vr is a matrix of order n x r.

Since diagonal elements of D matrix are nonzero, D is invertible.


So we can form following matrix, called as Pseudo inverse of A,

A+ = Vr D-1Urt
Singular Value Decomposition Page 7
A+ = Vr D-1Urt

Singular Value Decomposition Page 8

You might also like