AMPL parameters

The AMPL parameter description is very concise, but you can use the extensive parameter documentation in the User Manual as a reference, if you convert from WORHP's naming convention SomeParameter to AMPL's some_parameter. Note that not all NLP and QP parameters are available through the AMPL interface.

AMPL parameter list

accept_tol_feas        real: Tolerance for acceptable feasibility
accept_tol_opti        real: Tolerance for acceptable optimality
alpha_min_const        bool: Use a constant lower bound on Armijo stepsize in Filter
armijo_beta            real: Trial stepsize decrease factor for Armijo rule
armijo_max_alpha       real: Initial alpha for Armijo rule
armijo_min_alpha       real: Lower bound on alpha for Armijo rule
armijo_min_alpha_rec   real: Lower bound on alpha for Armijo rule during recovery
armijo_sigma           real: Scale factor for linearised descent check in Armijo rule
auto_qp_recovery       bool: Enable automatic QP recovery
betts_factor           real: Update factor for Betts' Hessian regularisation
betts_point            real: Smallest eigenvalue of the regularised Hessian
bfgs_max_blocksize     int : Maximum BFGS block size (depends on BFGS method)
bfgs_method            int : Choose BFGS method (dense, block, sparse)
bfgs_min_blocksize     int : Minimum BFGS block size (depends on BFGS method)
bfgs_restart           int : Restart BFGS update after this many iterations
bound_tol_fac          real: Factor in determining active constraints by KKT
check_fj               real: Upper bound used by Fritz-John heuristic
curv_b_cond            real: Block BFGS curvature condition bound
curv_b_fac             real: Block BFGS curvature condition regularisation factor
curv_cond              real: BFGS Curvature condition bound
curv_fac               real: BFGS curvature condition regularisation factor
cut_length             real: Scaling factor for Cut recovery strategy
eps                    real: Machine epsilon
feasible_dual          bool: Activate dual feasibility mode
feasible_init          bool: Activate initial feasibility mode
feasible_init_tol      real: Feasibility tolerance for no-objective feasible mode
feasible_only          bool: Activate feasible-only mode
fg_together            bool: F and G cannot be evaluated separately
fidif_eps              real: Finite difference perturbation
fidif_hm               bool: Approximate Hessian by finite differences (otherwise BFGS)
filter_bisec_alpha     bool: Filter heuristic to save Armijo iterations
filter_gamma_cv        real: Constraint violation decrease factor in Filter acceptance check
filter_gamma_f         real: Objective decrease factor in Filter acceptance check
filter_intersec_alpha  bool: Filter heuristic to save Armijo iterations
first_dif_central      bool: Use central finite difference quotient for first derivatives
fj_and_nd              bool: Enable Fritz-John and non-differentiable check heuristics
focus_on_feas          bool: Enable Focus-on-Feasibility mode
focus_on_feas_factor   real: Factor in Focus-on-Feasibility mode
gamma_alpha            real: Safety factor for alphamin calculation by Filter
ignore_filter_crit     bool: Activate accelerating heuristics for Filter
inc_betts_tau          real: Increase factor for Betts' update dampening term
inc_betts_tau_more     real: Larger increase factor for Betts' update dampening term
increase_iws           real: Increase factor for estimated integer workspace requirement
increase_rws           real: Increase factor for estimated real workspace requirement
infty                  real: Upper bound for numbers to be regarded as finite
infty_unbounded        real: Tolerance for unboundedness detection heuristic
initial_lm_est         bool: Enable initial Lagrange multiplier estimate
keep_acceptable_sol    bool: Save acceptable solutions as fallback
lin_mult               bool: Control Lagrange multiplier update
lm_est_qp_ip_com_tol   real: IP complementarity tolerance in initial multiplier estimate
lm_est_qp_ip_res_tol   real: IP residual tolerance in initial multiplier estimate
log_level              int : Enable XML logfiles and writing interval
log_result             int : Enable XML result logging and detail level
low_pass_alpha_f       real: Lowpass-filter update factor for objective values
low_pass_alpha_g       real: Lowpass-filter update factor for constraint values
low_pass_alpha_merit   real: Lowpass-filter update factor for merit function values
low_pass_filter        bool: Enable lowpass-filter termination criterion
ma57_pivot_thresh      real: Pivoting tolerance for MA57 = CNTL(1)
max_calls              int : Upper bound to Reverse Communication calls
max_force              int : Maximum number of Force recovery strategy steps
max_iter               int : Upper bound on major iterations
max_ls_counter         int : Control activation of Filter acceleration heuristics
max_norm               bool: Select max-norm instead of 1-norm in Filter
maxit                  int : Maximum number of major iterations (same as max_iter)
merit_function         int : Select merit function and penalty update [0, 3..5]
merit_grad_tol         real: Threshold of meritfunction gradient for increasing Hessian regularisation
min_betts_tau          real: Lower bound for Betts' update dampening term
more_relax             bool: Introduce one relaxation variable for every constraint
nlp_method             int : Select (1) Meritfunction or (3) Filter globalisation
nlp_print              int : NLP print level [-1..4]
outlev                 int : Solver verbosity [-1...4] (same as nlp_print)
qp.ls_method           int : Select linear solver for IP method
qp.max_iter            int : Maximum number of minor iterations
qp.print               int : QP solver verbosity [0..2]
qp.tol_comp            real: Complementarity tolerance for IP method
qp.tol_ls              real: Direct linear solver tolerance
qp.tol_res             real: Residual tolerance for IP method
qp_scale_param         real: (currently unused)
reduce_betts_tau       real: Decrease factor for Betts' update dampening term
reg_strategy           int : Select Hessian regularisation strategy in Filter
reinit_filter          bool: Enables Filter-reinitialisation accelerating heuristic
relax_max_delta        real: Upper bound for accepting the constraint relaxation variable
relax_max_pen          real: Upper bound on the constraint relaxation penalty
relax_rho              real: Update factor for the constraint relaxation penalty
relax_start            real: Initial value of the constraint relaxation penalty
rest_until_feas        bool: Do restoration until a feasible solution is found
scale_con_iter         bool: Scale constraints in every iteration
scale_fac_obj          real: Value to scale large objective functions to
scale_fac_qp           real: Upper bound on resulting matrix norm for QP scaling
scaled_fd              bool: Use a scaled perturbation for finite differences
scaled_kkt             bool: Scale KKT conditions
scaled_obj             bool: Scale the objective function
scaled_qp              bool: Scale some matrices handed to the QP
start_betts_tau        real: Initial value for Betts' update dampening term
switching_delta        real: Filter switching condition parameter
switching_scv          real: Filter switching condition parameter
switching_sf           real: Filter switching condition parameter
take_qp_sol            bool: Evaluate QP search direction regardless of convergence
timeout                real: Timeout in seconds
tol_comp               real: Complementarity tolerance
tol_feas               real: Feasibility tolerance
tol_opti               real: Optimality tolerance
tol_weak_active        real: (experimental)
too_big                bool: Enable too-big termination heuristics
too_big_cv             real: Upper bound on constraint violation for too-big heuristic
too_big_kkt            real: Upper bound on KKT values for too-big heuristic
user_df                bool: Objective gradient values supplied by caller
user_dg                bool: Jacobian values supplied by caller
user_hm                bool: Hessian values supplied by caller
version                report version
wantsol                solution report without -AMPL: sum of
		1 ==> write .sol file
		2 ==> print primal variable values
		4 ==> print dual variable values
		8 ==> do not print solution message
weak_active_set        bool: (experimental)