| -qps_mpgp_alpha_direct | - true sets expansion step length value directly, false (default) multiplier (typical between (0,2]) for step length equal to reciprocal of maximal eigenvalue
|
| -qps_mpgp_alpha | - fixed step length value for expansion, default: 2.0
|
| -qps_mpgp_gamma | - proportionality constant
|
| -qps_mpgp_maxeig | - approximate maximum eigenvalue of the Hessian (automatically computed by default)
|
| -qps_mpgp_maxeig_tol | - relative tolerance for power method to find approximate maximum eigenvalue of the Hessian
|
| -qps_mpgp_maxeig_iter | - number of iterations of power method to find an approximate maximum eigenvalue of the Hessian
|
| -qps_mpgp_btol | - boundary overshoot tolerance, default: 10*PETSC_MACHINE_EPSILON"
|
| -qps_mpgp_bound_chop_tol | - sets boundary to 0 for |boundary|<tol, default: 0
|
| -qps_mpgp_expansion_type | - set expansion step type, default: "std"
|
| -qps_mpgp_expansion_length_type | - set expansion step length type, default: "fixed"
|
| -qps_mpgp_alpha_reset | - if alpha=Nan reset to initial value, otherwise keep last alpaha, default: true
|
| -qps_mpgp_fallback | - throw away expansion step if cost function increased and do a std expansion step, default false
|
| -qps_mpgp_fallback2 | - same as fallback which is done only if the next step is proportioning
|