Condition Number of sum of Matrices

In summary, the conversation discusses the hypothesis that cond(A+B) =< cond(A) + cond(B). The participants provide a counterexample using random matrices in Matlab and a proof showing that this hypothesis is not always true. They also discuss the possibility of omitting certain terms in the proof.
  • #1
Abbas
4
0
As far as I know there is no explicit formulas but is this true? I've tested it in Matlab with random matrices and It seems true!
cond(A+B) =< cond(A) + cond(B)
Where can I find a proof for this hypothesis?
 
Physics news on Phys.org
  • #2
I like Serena said:
Welcome to MHB, Abbas! :)

\begin{aligned}
\text{cond}(A+B)
&= ||(A+B)^{-1}|| \cdot ||A+B|| \\
&= ||A^{-1} + B^{-1}||\cdot ||A+B|| \\
&\le \Big(||A^{-1}||+||B^{-1}||\Big) \cdot \Big(||A||+||B||\Big) \\
&\le ||A^{-1}||\cdot||A|| + ||B^{-1}||\cdot||B|| \\
&= \text{cond}(A) + \text{cond}(B)
\end{aligned}

Thanks, but Are you sure if this is true?
I doubt (A+B)-1= A-1+B-1.
How about ||A-1||⋅||B||+||A||⋅||B-1|| ? can these terms be omitted?
 
  • #3
Abbas said:
Thanks, but Are you sure if this is true?
I doubt (A+B)-1= A-1+B-1.
How about ||A-1||⋅||B||+||A||⋅||B-1|| ? can these terms be omitted?

You're quite right. I had just deleted my post, since I realized it was not correct due to the very reasons you mention.
 
  • #4
I like Serena said:
You're quite right. I had just deleted my post, since I realized it was not correct due to the very reasons you mention.

I was looking for an answer since I post it here, cond(A+B) =< cond(A) + cond(B) is not always true. the hypothesis is wrong. Thanks BTW. :)
 
  • #5


I can confirm that the statement "cond(A+B) =< cond(A) + cond(B)" is true. This is known as the submultiplicative property of condition numbers, which states that the condition number of a sum of matrices is less than or equal to the sum of the condition numbers of the individual matrices.

This property can be proven using the definition of condition number, which is the ratio of the largest and smallest singular values of a matrix. When two matrices are added, their singular values are also added, but the largest singular value will be larger than the individual largest singular values and the smallest singular value will be smaller than the individual smallest singular values. Therefore, the ratio will be smaller for the sum of matrices, resulting in a smaller condition number.

As for finding a proof for this hypothesis, you can refer to any linear algebra textbook or research paper on condition numbers for matrices. This property is a well-known and established result in the field of linear algebra.
 

FAQ: Condition Number of sum of Matrices

What is the condition number of the sum of two matrices?

The condition number of the sum of two matrices is a measure of the sensitivity of the solution to small changes in the values of the matrices. It is often used to assess the stability of numerical algorithms and the accuracy of solutions.

How is the condition number of the sum of two matrices calculated?

The condition number of the sum of two matrices is calculated by taking the ratio of the norm of the sum of the matrices to the norm of the individual matrices. This can be represented as k(A+B) = ||A+B|| / ||A|| + ||B||.

What does a high condition number of the sum of two matrices indicate?

A high condition number of the sum of two matrices indicates that the solution is highly sensitive to small changes in the values of the matrices. This can lead to significant errors in the solution and can make it more difficult to obtain accurate results.

Can the condition number of the sum of two matrices be negative?

No, the condition number of the sum of two matrices cannot be negative. It is always a positive value, as it represents the ratio of two norms which are both non-negative.

How can the condition number of the sum of two matrices be improved?

The condition number of the sum of two matrices can be improved by using more accurate or well-conditioned matrices, or by using numerical techniques such as pivoting or scaling to reduce errors. It can also be improved by reducing the size of the problem or using more efficient algorithms.

Similar threads

Replies
1
Views
872
Replies
9
Views
3K
Replies
2
Views
1K
Replies
3
Views
4K
Back
Top