1

I am currently using MATLAB to do matrix division of very large, very sparse, complex matrices that are symmetric in structure, but asymmetric in value (i.e. A(1,2)=3+4i and A(2,1)=3-4i).

I am now converting my code to Java. What is the proper equivalent function of A\b in Suitesparse/LApack?

I know this is what MATLAB is running for A\b, but chol seems to be limited to real, symmetric matrices.

FlyingTurbine
  • 323
  • 2
  • 9
  • You can not find an equivalent in speed. Maybe you can find a SuitSparse wrapper in Java... underneath Matlab code is elegant in C – Aznaveh Oct 28 '16 at 18:54
  • Welcome to Stackoverflow! Since the sample terms are conjugate and since I see the keyword `chol` in your question, I wonder if your matrix is a Hermitian, positive-definite matrix. If it is the case, Cholesky decomposition can be computed, even if the matrix is complex. Moreover, the Petsc library can provide you with interfaces to different Cholesky solvers listed [here](https://www.mcs.anl.gov/petsc/documentation/linearsolvertable.html) (direct solver/Cholesky). For complex matrices (X in last column), there is the one of Petsc and CHOLMOD of SuiteSparse, but also MUMPS and PaStiX (INRIA). – francis Oct 28 '16 at 20:32
  • @francis (I hope I used the notify feature correctly), I know that the real and imaginary portions of my matrix are Hermitian if taken separately (e.g. in MATLAB, `ishermitian(real(A))` and `ishermitian(imag(A))` are both true). – FlyingTurbine Oct 29 '16 at 01:36

0 Answers0