[petsc-users] Initial estimation on SNES and KSP
Dear all SNES uses internally a KSP to solve the linear system of equations right? Now the case that we had a linear system of equations that we are solving with SNES, how could we set the initial estimation for the KSP? If we just included the option -ksp_initial_guess_nonzero, the KSP will grab the vector X we passed to the SNES? Thanks in advance. Miguel -- *Miguel Angel Salazar de Troya* Graduate Research Assistant Department of Mechanical Science and Engineering University of Illinois at Urbana-Champaign (217) 550-2360 salaz...@illinois.edu
Re: [petsc-users] Initial estimation on SNES and KSP
Miguel Angel Salazar de Troya salazardetr...@gmail.com writes: Dear all SNES uses internally a KSP to solve the linear system of equations right? Now the case that we had a linear system of equations that we are solving with SNES, how could we set the initial estimation for the KSP? If we just included the option -ksp_initial_guess_nonzero, the KSP will grab the vector X we passed to the SNES? You definitely don't want this for Newton-type methods. If you have a nonzero guess for the solution of the linear system, you should have evaluated the Jacobian at that point. A zero initial guess is optimal for Newton-type methods. pgpz7SreS0X9D.pgp Description: PGP signature
Re: [petsc-users] Initial estimation on SNES and KSP
On Thu, Sep 4, 2014 at 12:28 PM, Jed Brown j...@jedbrown.org wrote: Miguel Angel Salazar de Troya salazardetr...@gmail.com writes: Dear all SNES uses internally a KSP to solve the linear system of equations right? Now the case that we had a linear system of equations that we are solving with SNES, how could we set the initial estimation for the KSP? If we just included the option -ksp_initial_guess_nonzero, the KSP will grab the vector X we passed to the SNES? You definitely don't want this for Newton-type methods. If you have a nonzero guess for the solution of the linear system, you should have evaluated the Jacobian at that point. A zero initial guess is optimal for Newton-type methods. Notice that Newton is solving for the correction, not the solution itself. Matt -- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener
