Differing definitions of Matrix Condition Number

478 Views Asked by At

When describing the condition number of a correlation matrix I have seen is described as the ratio of the singular values (from singluar value decomposition).

I have also seen it described as the ratio of the eigenvalues (from PCA analysis).

Can both be correct (under say different norms)? If so which one would be considered standard?

1

There are 1 best solutions below

2
On

For a symmetric positive semi-definite matrix the definitions are equivalent since singular values are eigenvalues.

The computation of eigenvalues is expensive, so often a condition number is defined as norm of matrix $A$ times the corresponding norm of the inverse. Thus a wide variety of definitions may be encountered. The operator norm of a matrix induced by Euclidean (sum-of-squares) vector norms will give the condition number defined as above, a ratio of largest to smallest singular values.