Singular matrix and MATLAB inversion

54 vues (au cours des 30 derniers jours)
Mehdi Chahine AMROUCHE
Mehdi Chahine AMROUCHE le 11 Juin 2019
Commenté : Bruno Luong le 26 Sep 2020
How come that det(A) = 0, and yet MATLAB computes the inverse of A (i.e inv(A) is computed without any warnings).
Where A is a square symmetric matrix.
Thanks !
  3 commentaires
Shivam Sharma
Shivam Sharma le 26 Sep 2020
let's say A = [1 2 3; 4 5 6; 7 8 9]
Bruno Luong
Bruno Luong le 26 Sep 2020
@Sharma I get the warning wth your matrix, so it's not a valide example
>> A = [1 2 3; 4 5 6; 7 8 9]:
>> inv(A)
Warning: Matrix is close to singular or badly scaled. Results may be inaccurate. RCOND = 2.202823e-18.
ans =
1.0e+16 *
0.3153 -0.6305 0.3153
-0.6305 1.2610 -0.6305
0.3153 -0.6305 0.3153

Connectez-vous pour commenter.

Réponse acceptée

Steven Lord
Steven Lord le 11 Juin 2019
DON'T use det to determine if a matrix is singular!
A non-zero multiple of the identity matrix isn't singular, right?
A = 0.1*eye(400);
det(A)
The determinant of A underflowed to 0.
Can multiplying a singular (or nearly singular) matrix by a non-zero scalar make it no longer singular?
B = [1 1; 1 1+eps];
C = 1e8*B;
det(B)
det(C)
A number with a determinant on the order of 1 can't be singular, can it?
Use the cond or rcond functions instead. Matrices with condition numbers closer to 1 are better behaved.
cond(A)
cond(B)
cond(C)
If you're calling inv to try to solve a system of linear equations, DON'T. Use the backslash operator (\) instead. Even if your textbook told you to multiply by the inverse matrix, well, in theory there is no difference between theory and practice ...
  1 commentaire
Bjorn Gustavsson
Bjorn Gustavsson le 12 Juin 2019
If you have to solve a system of equations where the matrix is singular or ill-conditioned don't use backslash either. To "give" a solution in that case one should know what has been done and what has not been done. The best way to do that is to SVD-based Moore-Penrose inverse, with suitable (zeroth- second-order Tikhonov ) regularization, inspection of the singular-values, and the model-space eigenvectors etc. A great utility for this is:

Connectez-vous pour commenter.

Plus de réponses (1)

Bjorn Gustavsson
Bjorn Gustavsson le 11 Juin 2019
Are you absolutely sure you haven't turned off the warnings, in your startup.m or elsewhere?

Catégories

En savoir plus sur Linear Algebra dans Help Center et File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by