Hi,

I just have some questions using Eigen Linear Algebra fonctions.

1- Is it possible to use FullPivLU on a invertible Matrix returning
P^(-1)*L*U without the permutationQ ?
Actually I just don't want to use PartialPivLU so as to get better
results (accuracy) (for exemple ill conditioned Matrix)

2- Resolving : ( A.X = B ) If A is invertible what's the difference by using
X=A.lu().solve(B)
X=A.fullPivLu().solve(B)

I also tried :

X=A.inverse()*B --> It doesn't seems to be any difference time
calculation (even for dense Matrix)

to solve the system.

Thks for answers.

theokoles wrote:Hi,

I just have some questions using Eigen Linear Algebra fonctions.

1- Is it possible to use FullPivLU on a invertible Matrix returning
P^(-1)*L*U without the permutationQ ?
Actually I just don't want to use PartialPivLU so as to get better
results (accuracy) (for exemple ill conditioned Matrix)

hm, well, if you want full pivoting LU with Q=I, then you end up with a partial pivoting LU.

2- Resolving : ( A.X = B ) If A is invertible what's the difference by using
X=A.lu().solve(B)
X=A.fullPivLu().solve(B)

I also tried :

X=A.inverse()*B --> It doesn't seems to be any difference time
calculation (even for dense Matrix)

to solve the system.

X=A.fullPivLu().solve(B) : most accurate but slower of the three
X=A.lu().solve(B) : accurate enough 99% of the time and the fastest of the three
X=A.inverse()*B : less accurate and more expensive than X=A.lu().solve(B)

thks ^^