[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
## Re: inverse Matrix problem, Newton's method

**From**: |
Sebastian Schöps |

**Subject**: |
Re: inverse Matrix problem, Newton's method |

**Date**: |
Sat, 10 Nov 2018 16:10:08 -0600 (CST) |

mecht wrote
>* My problem is that i can not produce any result due to false(?) matrix*
>* singularity. I tried left-hand matrix division, LU method, inv(A) but*
>* Octave warns about matrix singularity due to machine precision.*
>* Octave returns 0 or extremely low value (like 1.0E-100) when asked for*
>* det(A)*
Are you sure that your boundary conditions are correct? In particular, may
there a Dirichlet condition be missing? A missing boundary condition is the
most common reason for a FEM matrix to be singular. Does it work for linear
problems?
In either case, do not use "inv", "rank" or "det" to solve, check the
regularity or rank of your matrix. Those algorithms are not reliable for
large problems. Just use Gauss or lu but make sure to use a sparse matrix
decomposition. Indeed, the most convenient choice is using the backslash
("x=A\b"). It will automatically take the best approach. If it fails then
almost surely you have made a mistake...
Sebastian
--
Sent from: http://octave.1599824.n4.nabble.com/Octave-General-f1599825.html