Scribd
Upload
696 views4 pages

The Cholesky Decomposition

The document discusses the Cholesky decomposition method for solving systems of linear equations. It can be faster than LU decomposition by exploiting the property of symmetric matrices. However, the matrix must be Hermitian and positive definite. This means the matrix A is decomposed as A=LLT, where L is a lower triangular matrix with positive diagonals. The document then provides an example using NumPy to calculate the Cholesky decomposition of a sample matrix and solve for the variable values.

Uploaded by

hachan
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
696 views4 pages

The Cholesky Decomposition

The document discusses the Cholesky decomposition method for solving systems of linear equations. It can be faster than LU decomposition by exploiting the property of symmetric matrices. However, the matrix must be Hermitian and positive definite. This means the matrix A is decomposed as A=LLT, where L is a lower triangular matrix with positive diagonals. The document then provides an example using NumPy to calculate the Cholesky decomposition of a sample matrix and solve for the variable values.

Uploaded by

hachan
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/ 4

The Cholesky decomposition

The Cholesky decomposition is another way of solving systems of linear equations. It can
be significantly faster and uses a lot less memory than the LU decomposition, by exploiting
the property of symmetric matrices. However, the matrix being decomposed must be
Hermitian (or real-valued symmetric and thus square) and positive definite. This means
that the Amatrix is decomposed as A=LLT, where L is a lower triangular matrix with real and
positive numbers on the diagonals, and LT is the conjugate transpose of L.

Let's consider another example of a system of linear equations where the Amatrix is both
Hermitian and positive definite. Again, the equation is in the form of Ax=B,
where A and B take the following values:
Let's represent these matrices as NumPy arrays:

In [ ]:
"""
Cholesky decomposition with NumPy
"""
import numpy as np

A = np.array([
[10., -1., 2., 0.],
[-1., 11., -1., 3.],
[2., -1., 10., -1.],
[0., 3., -1., 8.]])
B = np.array([6., 25., -11., 15.])

L = np.linalg.cholesky(A)

The cholesky() function of numpy.linalg would compute the lower triangular factor of
the A matrix. Let's view the lower triangular matrix:

In [ ]:
In [ ]:
print(L)
Out[ ]:
[[ 3.16227766 0. 0. 0. ]
[-0.31622777 3.3015148 0. 0. ]
[ 0.63245553 -0.24231301 3.08889696 0. ]
[ 0. 0.9086738 -0.25245792 2.6665665 ]]

To verify that the Cholesky decomposition results are correct, we can use the definition of
the Cholesky factorization by multiplying L by its conjugate transpose, which will lead us
back to the values of the A matrix:

In [ ]:
print(np.dot(L, L.T.conj())) # A=L.L*
Out [ ]:
[[10. -1. 2. 0.]
[-1. 11. -1. 3.]
[ 2. -1. 10. -1.]
[ 0. 3. -1. 8.]]

Before solving for x, we need to solve for LTx as y. Let's use the solve() method of numpy.linalg:

In [ ]:
y = np.linalg.solve(L, B) # L.L*.x=B; When L*.x=y, then L.y=B

To solve for x, we need to solve again using the conjugate transpose of L andy:

In [ ]:
x = np.linalg.solve(L.T.conj(), y) # x=L*'.y

Let's print our result of x:


In [ ]:
print(x)
Out[ ]:
[ 1. 2. -1. 1.]

The output gives us our values of x for a, b, c, and d.

To show that the Cholesky factorization gives us the correct values, we can verify the answer
by multiplying the A matrix by the transpose of x to return the values of B:

In [ ] :
print(np.mat(A) * np.mat(x).T) # B=Ax
Out[ ]:

[[ 6.]

[ 25.]

[-11.]

[ 15.]]

This shows that the values of x by the Cholesky decomposition would lead to the same values given by B.

You might also like