Options

This document lists the configuration options that Uno accepts, their types and default values. Options are grouped by functional area. Defaults are taken from uno/options/DefaultOptions.cpp.

Ingredients

Ingredient	Available strategies
constraint_relaxation_strategy	feasibility_restoration
inequality_handling_method	inequality_constrained, interior_point
hessian_model	exact, LBFGS, LSR1, identity, zero
inertia_correction_strategy	primal, primal_dual, none
globalization_mechanism	TR, LS
globalization_strategy	merit_function, fletcher_filter_method, waechter_filter_method, funnel_method

Subproblem solver

Ingredient	Available strategies
QP_solver	BQPD, HiGHS (if available)
LP_solver	BQPD, HiGHS (if available)
linear_solver	MA57, MA27, MA86, MUMPS, SSIDS (if available)

If not provided, the solver is chosen automatically from the available solvers (if any).

Option	Type	Default	Description
libhsl_path	string	""	Name/path of the HSL shared library to dlopen for MA27/MA57/MA86 (only used when Uno is built with -DHSL_RUNTIME_LOADING=ON). An empty value resolves at runtime to the platform default libhsl.{so,dylib,dll}. See the installation guide for the resolution order and the UNO_HSL_LIBRARY environment variable.

Termination

Option	Type	Default	Description
primal_tolerance	double	1e-8	Tolerance on constraint violation
dual_tolerance	double	1e-8	Tolerance on stationarity and complementarity
loose_primal_tolerance	double	1e-6	Loose tolerance on constraint violation
loose_dual_tolerance	double	1e-6	Loose tolerance on stationarity and complementarity
loose_tolerance_iteration_threshold	integer	15	Number of iterations for the loose tolerance to apply
max_iterations	integer	2000	Maximum number of outer iterations
time_limit	double	infinity	Time limit
print_solution	bool	false	Whether the primal-dual solution is printed
unbounded_objective_threshold	double	-1e20	Objective threshold under which the problem is declared unbounded

Main options

String options

Option	Possible values	Default	Description
preset	filtersqp, ipopt, auto	auto	Uno preset
logger	SILENT, DISCRETE, WARNING, INFO, DEBUG, DEBUG2, DEBUG3	INFO	Verbosity level of the logger
progress_norm	L1, L2, INF	L1	Norm used for the progress measures
residual_norm	L1, L2, INF	INF	Norm used for the residuals

Numerical options

Option	Type	Default	Description
residual_scaling_threshold	double	100.0	Scaling factor in stationarity and complementarity residuals
protect_actual_reduction_against_roundoff	bool	false	Whether the actual reduction is slightly modified to account for roundoff
protected_actual_reduction_macheps_coefficient	double	10	Coefficient of the machine epsilon in the protected actual reduction
print_subproblem	bool	false	Whether the subproblem is printed in DEBUG mode
write_solution_to_file	bool	false	Whether the solution is printed to a file (used by AMPL and CUTEst)

Globalization strategy options

Option	Type	Default	Description
armijo_decrease_fraction	double	1e-4	Fraction of the predicted reduction that should be achieved by the actual reduction in the Armijo condition
armijo_tolerance	double	1e-9	Minimum value of the predicted reduction in the Armijo condition, 0 otherwise

Switching method options

Option	Type	Default	Description
switching_delta	double	0.999	Fraction of the constraint violation that should be achieved by the predicted reduction in the switching condition
switching_infeasibility_exponent	double	2	Exponent of the constraint violation in the switching condition

Merit function options

Option	Type	Default	Description
sufficient_infeasibility_decrease_ratio	double	0.9	Sufficient infeasibility decrease ratio

Filter method options

String options

Option	Possible values	Default	Description
filter_type	standard	standard	Type of the filter data structure

Numerical options

Option	Type	Default	Description
filter_beta	double	0.999	Fraction in the infeasibility sufficient reduction condition
filter_gamma	double	0.001	Slope in the objective sufficient reduction condition
filter_ubd	double	1e2	Minimum value for the initial upper bound on the infeasibility
filter_fact	double	1.25	Multiple of the initial infeasibility for the initial upper bound on the infeasibility
filter_capacity	integer	50	Maximum number of filter entries
filter_sufficient_infeasibility_decrease_factor	double	0.9	Infeasibility decrease factor in the infeasibility sufficient decrease condition
filter_reset_iteration_threshold	integer	5	Number of consecutive filter rejections before the filter is reset
max_number_filter_resets	integer	5	Maximum number of filter resets

Funnel options

Option	Type	Default	Description
funnel_kappa	double	0.5	Convex combination coefficient in funnel update rule
funnel_beta	double	0.9999	Fraction in the infeasibility sufficient reduction condition
funnel_gamma	double	0.001	Slope in the objective sufficient reduction condition
funnel_ubd	double	1.0	Minimum value for the initial upper bound on the infeasibility
funnel_fact	double	1.5	Multiple of the initial infeasibility for the initial upper bound on the infeasibility
funnel_update_strategy	integer	1	Rule for the funnel update (1, 2, or 3)
funnel_require_acceptance_wrt_current_iterate	bool	false	Whether the trial iterate should improve upon the current iterate

Line search options

Option	Type	Default	Description
LS_backtracking_ratio	double	0.5	Decrease ratio of the step length for backtracking line search
LS_min_step_length	double	1e-12	Minimum acceptable step length before failure is reported
LS_scale_duals_with_step_length	bool	true	Whether the Lagrange multipliers are scaled with the step length

Second-Order Corrections (SOC) options

Option	Type	Default	Description
SOC_max_iterations_SOC	integer	4	Maximum number of SOC iterations
SOC_infeasibility_fraction	double	0.99	Fraction of the infeasibility in the SOC termination criterion

Inertia correction options

Option	Type	Default	Description
regularization_failure_threshold	double	1e20	Threshold for the primal inertia correction coefficient above which failure is reported
primal_regularization_initial_factor	double	1e-4	Initial value of primal inertia correction coefficient
regularization_increase_factor	double	2	Increase factor for the primal inertia correction coefficient
dual_regularization_fraction	double	1e-8	Fraction of the dual inertia correction parameter
primal_regularization_lb	double	1e-20	Minimum value of the primal inertia correction coefficient upon decrease
primal_regularization_decrease_factor	double	3.0	Decrease factor for the primal inertia correction coefficient
primal_regularization_fast_increase_factor	double	100.0	Fast increase factor for the primal inertia correction coefficient
primal_regularization_slow_increase_factor	double	8.0	Slow increase factor for the primal inertia correction coefficient
threshold_unsuccessful_attempts	integer	8	Number of unsuccessful attempts until inertia correction becomes more aggressive

Quasi-Newton options

Option	Type	Default	Description
quasi_newton_memory_size	integer	6	Size of the quasi-Newton limited memory
quasi_newton_delta_lower_bound	double	1e-8	Lower bound on delta in quasi-Newton approximations
quasi_newton_delta_upper_bound	double	1e8	Upper bound on delta in quasi-Newton approximations
LBFGS_max_skips_before_reset	integer	2	Number of consecutive skips before the limited memory is reset
LSR1_pivot_max_magnitude	double	1e-7	Maximum magnitude of allowed pivots in L-SR1

Trust region options

Option	Type	Default	Description
TR_radius	double	10.0	Initial value of the radius
TR_increase_factor	double	2	Increase factor of the radius for successful iterations
TR_decrease_factor	double	2	Decrease factor of the radius for unsuccessful iterations
TR_aggressive_decrease_factor	double	4	Decrease factor of the radius when errors occur
TR_activity_tolerance	double	1e-6	Tolerance with which the trust-region constraint is considered active
TR_min_radius	double	1e-12	Minimum radius acceptable before failure is reported
TR_radius_reset_threshold	double	1e-4	Smallest value to which the radius is reset at the beginning of an iteration

Feasibility restoration options

Option	Type	Default	Description
switch_to_optimality_requires_linearized_feasibility	bool	true	Whether the switch to optimality phase requires the linearized constraints to be consistent
l1_constraint_violation_coefficient	double	1	Coefficient of the constraint violation in the $\ell_1$ relaxed problem

Barrier subproblem options

String options

Option	Possible values	Default	Description
barrier_function	log	log	Type of the barrier function

Numerical options

Option	Type	Default	Description
barrier_initial_parameter	double	0.1	Initial value of the barrier parameter
barrier_default_multiplier	double	1	Initial value of the bound multipliers
barrier_tau_min	double	0.99	Coefficient of the fraction-to-boundary rule
barrier_k_sigma	double	1e10	Safeguard parameter for rescaling the bound multipliers
barrier_k_mu	double	0.2	Coefficient for multiplicative update of the barrier parameter
barrier_theta_mu	double	1.5	Coefficient for geometric update of the barrier parameter
barrier_k_epsilon	double	10	Scaling factor of the barrier parameter in the barrier update rule
barrier_update_fraction	double	10	Fraction in the barrier update rule
barrier_regularization_exponent	double	0.25	Exponent of the barrier parameter in the dual inertia correction parameter
barrier_small_direction_factor	double	10.0	Multiple of the machine epsilon in the small step condition
barrier_push_variable_to_interior_k1	double	1e-2	Coefficient for the perturbation of the initial bounds
barrier_push_variable_to_interior_k2	double	1e-2	Coefficient for the perturbation of the initial bounds
barrier_damping_factor	double	1e-5	Damping coefficient for single bounds
barrier_small_infeasibility_factor	double	1e-4	Factor in the small infeasibility test
least_square_multiplier_max_norm	double	1e3	Maximum accepted norm of the least-square multipliers

BQPD options

Option	Possible values	Default	Description
BQPD_kmax_heuristic	filtersqp, minotaur	filtersqp	Heuristic used to pick upper bound on nullspace size (kmax)