Home > matpower5.1 > ipopt_options.m

ipopt_options

PURPOSE ^

IPOPT_OPTIONS Sets options for IPOPT.

SYNOPSIS ^

function opt = ipopt_options(overrides, mpopt)

DESCRIPTION ^

IPOPT_OPTIONS  Sets options for IPOPT.

   OPT = IPOPT_OPTIONS
   OPT = IPOPT_OPTIONS(OVERRIDES)
   OPT = IPOPT_OPTIONS(OVERRIDES, FNAME)
   OPT = IPOPT_OPTIONS(OVERRIDES, MPOPT)

   Sets the values for the options.ipopt struct normally passed to
   IPOPT.

   Inputs are all optional, second argument must be either a string
   (FNAME) or a struct (MPOPT):

       OVERRIDES - struct containing values to override the defaults
       FNAME - name of user-supplied function called after default
           options are set to modify them. Calling syntax is:
                   MODIFIED_OPT = FNAME(DEFAULT_OPT);
       MPOPT - MATPOWER options struct, uses the following fields:
           opf.violation   - used to set opt.constr_viol_tol
           verbose         - used to opt.print_level
           ipopt.opts      - struct containing values to use as OVERRIDES
           ipopt.opt_fname - name of user-supplied function used as FNAME,
               except with calling syntax:
                   MODIFIED_OPT = FNAME(DEFAULT_OPT, MPOPT);
           ipopt.opt       - numbered user option function, if and only if
               ipopt.opt_fname is empty and ipopt.opt is non-zero, the value
               of ipopt.opt_fname is generated by appending ipopt.opt to
               'ipopt_user_options_' (for backward compatibility with old
               MATPOWER option IPOPT_OPT).

   Output is an options.ipopt struct to pass to IPOPT.

   There are multiple ways of providing values to override the default
   options. Their precedence and order of application are as follows:

   With inputs OVERRIDES and FNAME
       1. FNAME is called
       2. OVERRIDES are applied
   With inputs OVERRIDES and MPOPT
       1. FNAME (from ipopt.opt_fname or ipopt.opt) is called
       2. ipopt.opts (if not empty) are applied
       3. OVERRIDES are applied

   Example:

   If ipopt.opt = 3, then after setting the default IPOPT options,
   IPOPT_OPTIONS will execute the following user-defined function
   to allow option overrides:

       opt = ipopt_user_options_3(opt, mpopt);

   The contents of ipopt_user_options_3.m, could be something like:

       function opt = ipopt_user_options_3(opt, mpopt)
       opt.nlp_scaling_method = 'none';
       opt.max_iter           = 500;
       opt.derivative_test    = 'first-order';

   See the options reference section in the IPOPT documentation for
   details on the available options.

       http://www.coin-or.org/Ipopt/documentation/

   See also IPOPT, MPOPTION.

CROSS-REFERENCE INFORMATION ^

This function calls: This function is called by:

SOURCE CODE ^

0001 function opt = ipopt_options(overrides, mpopt)
0002 %IPOPT_OPTIONS  Sets options for IPOPT.
0003 %
0004 %   OPT = IPOPT_OPTIONS
0005 %   OPT = IPOPT_OPTIONS(OVERRIDES)
0006 %   OPT = IPOPT_OPTIONS(OVERRIDES, FNAME)
0007 %   OPT = IPOPT_OPTIONS(OVERRIDES, MPOPT)
0008 %
0009 %   Sets the values for the options.ipopt struct normally passed to
0010 %   IPOPT.
0011 %
0012 %   Inputs are all optional, second argument must be either a string
0013 %   (FNAME) or a struct (MPOPT):
0014 %
0015 %       OVERRIDES - struct containing values to override the defaults
0016 %       FNAME - name of user-supplied function called after default
0017 %           options are set to modify them. Calling syntax is:
0018 %                   MODIFIED_OPT = FNAME(DEFAULT_OPT);
0019 %       MPOPT - MATPOWER options struct, uses the following fields:
0020 %           opf.violation   - used to set opt.constr_viol_tol
0021 %           verbose         - used to opt.print_level
0022 %           ipopt.opts      - struct containing values to use as OVERRIDES
0023 %           ipopt.opt_fname - name of user-supplied function used as FNAME,
0024 %               except with calling syntax:
0025 %                   MODIFIED_OPT = FNAME(DEFAULT_OPT, MPOPT);
0026 %           ipopt.opt       - numbered user option function, if and only if
0027 %               ipopt.opt_fname is empty and ipopt.opt is non-zero, the value
0028 %               of ipopt.opt_fname is generated by appending ipopt.opt to
0029 %               'ipopt_user_options_' (for backward compatibility with old
0030 %               MATPOWER option IPOPT_OPT).
0031 %
0032 %   Output is an options.ipopt struct to pass to IPOPT.
0033 %
0034 %   There are multiple ways of providing values to override the default
0035 %   options. Their precedence and order of application are as follows:
0036 %
0037 %   With inputs OVERRIDES and FNAME
0038 %       1. FNAME is called
0039 %       2. OVERRIDES are applied
0040 %   With inputs OVERRIDES and MPOPT
0041 %       1. FNAME (from ipopt.opt_fname or ipopt.opt) is called
0042 %       2. ipopt.opts (if not empty) are applied
0043 %       3. OVERRIDES are applied
0044 %
0045 %   Example:
0046 %
0047 %   If ipopt.opt = 3, then after setting the default IPOPT options,
0048 %   IPOPT_OPTIONS will execute the following user-defined function
0049 %   to allow option overrides:
0050 %
0051 %       opt = ipopt_user_options_3(opt, mpopt);
0052 %
0053 %   The contents of ipopt_user_options_3.m, could be something like:
0054 %
0055 %       function opt = ipopt_user_options_3(opt, mpopt)
0056 %       opt.nlp_scaling_method = 'none';
0057 %       opt.max_iter           = 500;
0058 %       opt.derivative_test    = 'first-order';
0059 %
0060 %   See the options reference section in the IPOPT documentation for
0061 %   details on the available options.
0062 %
0063 %       http://www.coin-or.org/Ipopt/documentation/
0064 %
0065 %   See also IPOPT, MPOPTION.
0066 
0067 %   MATPOWER
0068 %   Copyright (c) 2010-2015 by Power System Engineering Research Center (PSERC)
0069 %   by Ray Zimmerman, PSERC Cornell
0070 %
0071 %   $Id: ipopt_options.m 2644 2015-03-11 19:34:22Z ray $
0072 %
0073 %   This file is part of MATPOWER.
0074 %   Covered by the 3-clause BSD License (see LICENSE file for details).
0075 %   See http://www.pserc.cornell.edu/matpower/ for more info.
0076 
0077 %%-----  initialization and arg handling  -----
0078 %% defaults
0079 verbose = 2;
0080 fname   = '';
0081 
0082 %% second argument
0083 if nargin > 1 && ~isempty(mpopt)
0084     if ischar(mpopt)        %% 2nd arg is FNAME (string)
0085         fname = mpopt;
0086         have_mpopt = 0;
0087     else                    %% 2nd arg is MPOPT (MATPOWER options struct)
0088         have_mpopt = 1;
0089         verbose = mpopt.verbose;
0090         if isfield(mpopt.ipopt, 'opt_fname') && ~isempty(mpopt.ipopt.opt_fname)
0091             fname = mpopt.ipopt.opt_fname;
0092         elseif mpopt.ipopt.opt
0093             fname = sprintf('ipopt_user_options_%d', mpopt.ipopt.opt);
0094         end
0095     end
0096 else
0097     have_mpopt = 0;
0098 end
0099 
0100 %%-----  set default options for IPOPT  -----
0101 %% printing
0102 if verbose
0103     opt.print_level = min(12, verbose*2+1);
0104 else
0105     opt.print_level = 0;
0106 end
0107 
0108 %% convergence
0109 opt.tol             = 1e-8;                %% default 1e-8
0110 opt.max_iter        = 250;                  %% default 3000
0111 opt.dual_inf_tol    = 0.1;                  %% default 1
0112 if have_mpopt
0113     opt.constr_viol_tol = mpopt.opf.violation;                  %% default 1e-4
0114     opt.acceptable_constr_viol_tol = mpopt.opf.violation*100;   %% default 1e-2
0115 end
0116 opt.compl_inf_tol   = 1e-5;                 %% default 1e-4
0117 opt.acceptable_tol  = 1e-8;                 %% default 1e-6
0118 % opt.acceptable_iter = 15;                   %% default 15
0119 % opt.acceptable_dual_inf_tol     = 1e+10;    %% default 1e+10
0120 opt.acceptable_compl_inf_tol    = 1e-3;     %% default 1e-2
0121 % opt.acceptable_obj_change_tol   = 1e+20;    %% default 1e+20
0122 % opt.diverging_iterates_tol      = 1e+20;    %% default 1e+20
0123 
0124 %% NLP scaling
0125 % opt.nlp_scaling_method  = 'none';           %% default 'gradient-based'
0126 
0127 %% NLP
0128 % opt.fixed_variable_treatment    = 'make_constraint';    %% default 'make_parameter'
0129 % opt.honor_original_bounds       = 'no';                 %% default 'yes'
0130 % opt.check_derivatives_for_naninf = 'yes';               %% default 'no'
0131 
0132 %% initialization
0133 % opt.least_square_init_primal    = 'yes';        %% default 'no'
0134 % opt.least_square_init_duals     = 'yes';        %% default 'no'
0135 
0136 %% barrier parameter update
0137 opt.mu_strategy                 = 'adaptive';   %% default 'monotone'
0138 
0139 %% linear solver
0140 % opt.linear_solver   = 'ma27';
0141 % opt.linear_solver   = 'ma57';
0142 % opt.linear_solver   = 'pardiso';
0143 % opt.linear_solver   = 'wsmp';
0144 % opt.linear_solver   = 'mumps';          %% default 'mumps'
0145 % opt.linear_solver   = 'custom';
0146 % opt.linear_scaling_on_demand    = 'no'; %% default 'yes'
0147 
0148 %% step calculation
0149 % opt.mehrotra_algorithm      = 'yes';    %% default 'no'
0150 % opt.fast_step_computation   = 'yes';    %% default 'no'
0151 
0152 %% restoration phase
0153 % opt.expect_infeasible_problem   = 'yes';    %% default 'no'
0154 
0155 %% derivative checker
0156 % opt.derivative_test         = 'second-order';   %% default 'none'
0157 
0158 %% hessian approximation
0159 % opt.hessian_approximation   = 'limited-memory'; %% default 'exact'
0160 
0161 % ma57 options
0162 %opt.ma57_pre_alloc = 3;
0163 %opt.ma57_pivot_order = 4;
0164 
0165 %%-----  call user function to modify defaults  -----
0166 if ~isempty(fname)
0167     if have_mpopt
0168         opt = feval(fname, opt, mpopt);
0169     else
0170         opt = feval(fname, opt);
0171     end
0172 end
0173 
0174 %%-----  apply overrides  -----
0175 if have_mpopt && isfield(mpopt.ipopt, 'opts') && ~isempty(mpopt.ipopt.opts)
0176     opt = nested_struct_copy(opt, mpopt.ipopt.opts);
0177 end
0178 if nargin > 0 && ~isempty(overrides)
0179     opt = nested_struct_copy(opt, overrides);
0180 end
0181 
0182 
0183 %--------------------------  Options Documentation  --------------------------
0184 % (as printed by IPOPT 3.8)
0185 % ### Output ###
0186 %
0187 % print_level                            0 <= (          5) <= 12
0188 %    Output verbosity level.
0189 %      Sets the default verbosity level for console output. The larger this
0190 %      value the more detailed is the output.
0191 %
0192 % output_file                   ("")
0193 %    File name of desired output file (leave unset for no file output).
0194 %      NOTE: This option only works when read from the ipopt.opt options file!
0195 %      An output file with this name will be written (leave unset for no file
0196 %      output).  The verbosity level is by default set to "print_level", but can
0197 %      be overridden with "file_print_level".  The file name is changed to use
0198 %      only small letters.
0199 %    Possible values:
0200 %     - *                       [Any acceptable standard file name]
0201 %
0202 % file_print_level                       0 <= (          5) <= 12
0203 %    Verbosity level for output file.
0204 %      NOTE: This option only works when read from the ipopt.opt options file!
0205 %      Determines the verbosity level for the file specified by "output_file".
0206 %      By default it is the same as "print_level".
0207 %
0208 % print_user_options            ("no")
0209 %    Print all options set by the user.
0210 %      If selected, the algorithm will print the list of all options set by the
0211 %      user including their values and whether they have been used.  In some
0212 %      cases this information might be incorrect, due to the internal program
0213 %      flow.
0214 %    Possible values:
0215 %     - no                      [don't print options]
0216 %     - yes                     [print options]
0217 %
0218 % print_options_documentation   ("no")
0219 %    Switch to print all algorithmic options.
0220 %      If selected, the algorithm will print the list of all available
0221 %      algorithmic options with some documentation before solving the
0222 %      optimization problem.
0223 %    Possible values:
0224 %     - no                      [don't print list]
0225 %     - yes                     [print list]
0226 %
0227 % print_timing_statistics       ("no")
0228 %    Switch to print timing statistics.
0229 %      If selected, the program will print the CPU usage (user time) for
0230 %      selected tasks.
0231 %    Possible values:
0232 %     - no                      [don't print statistics]
0233 %     - yes                     [print all timing statistics]
0234 %
0235 % option_file_name              ("")
0236 %    File name of options file (to overwrite default).
0237 %      By default, the name of the Ipopt options file is "ipopt.opt" - or
0238 %      something else if specified in the IpoptApplication::Initialize call. If
0239 %      this option is set by SetStringValue BEFORE the options file is read, it
0240 %      specifies the name of the options file.  It does not make any sense to
0241 %      specify this option within the options file.
0242 %    Possible values:
0243 %     - *                       [Any acceptable standard file name]
0244 %
0245 % replace_bounds                ("no")
0246 %    Indicates if all variable bounds should be replaced by inequality
0247 %    constraints
0248 %      This option must be set for the inexact algorithm
0249 %    Possible values:
0250 %     - no                      [leave bounds on variables]
0251 %     - yes                     [replace variable bounds by inequality
0252 %                                constraints]
0253 %
0254 % skip_finalize_solution_call   ("no")
0255 %    Indicates if call to NLP::FinalizeSolution after optimization should be
0256 %    suppressed
0257 %      In some Ipopt applications, the user might want to call the
0258 %      FinalizeSolution method separately.  Setting this option to "yes" will
0259 %      cause the IpoptApplication object to suppress the default call to that
0260 %      method.
0261 %    Possible values:
0262 %     - no                      [call FinalizeSolution]
0263 %     - yes                     [do not call FinalizeSolution]
0264 %
0265 % print_info_string             ("no")
0266 %    Enables printing of additional info string at end of iteration output.
0267 %      This string contains some insider information about the current iteration.
0268 %    Possible values:
0269 %     - no                      [don't print string]
0270 %     - yes                     [print string at end of each iteration output]
0271 %
0272 %
0273 %
0274 % ### Convergence ###
0275 %
0276 % tol                                    0 <  (      1e-08) <  +inf
0277 %    Desired convergence tolerance (relative).
0278 %      Determines the convergence tolerance for the algorithm.  The algorithm
0279 %      terminates successfully, if the (scaled) NLP error becomes smaller than
0280 %      this value, and if the (absolute) criteria according to "dual_inf_tol",
0281 %      "primal_inf_tol", and "cmpl_inf_tol" are met.  (This is epsilon_tol in
0282 %      Eqn. (6) in implementation paper).  See also "acceptable_tol" as a second
0283 %      termination criterion.  Note, some other algorithmic features also use
0284 %      this quantity to determine thresholds etc.
0285 %
0286 % s_max                                  0 <  (        100) <  +inf
0287 %    Scaling threshold for the NLP error.
0288 %      (See paragraph after Eqn. (6) in the implementation paper.)
0289 %
0290 % max_iter                               0 <= (       3000) <  +inf
0291 %    Maximum number of iterations.
0292 %      The algorithm terminates with an error message if the number of
0293 %      iterations exceeded this number.
0294 %
0295 % max_cpu_time                           0 <  (      1e+06) <  +inf
0296 %    Maximum number of CPU seconds.
0297 %      A limit on CPU seconds that Ipopt can use to solve one problem.  If
0298 %      during the convergence check this limit is exceeded, Ipopt will terminate
0299 %      with a corresponding error message.
0300 %
0301 % dual_inf_tol                           0 <  (          1) <  +inf
0302 %    Desired threshold for the dual infeasibility.
0303 %      Absolute tolerance on the dual infeasibility. Successful termination
0304 %      requires that the max-norm of the (unscaled) dual infeasibility is less
0305 %      than this threshold.
0306 %
0307 % constr_viol_tol                        0 <  (     0.0001) <  +inf
0308 %    Desired threshold for the constraint violation.
0309 %      Absolute tolerance on the constraint violation. Successful termination
0310 %      requires that the max-norm of the (unscaled) constraint violation is less
0311 %      than this threshold.
0312 %
0313 % compl_inf_tol                          0 <  (     0.0001) <  +inf
0314 %    Desired threshold for the complementarity conditions.
0315 %      Absolute tolerance on the complementarity. Successful termination
0316 %      requires that the max-norm of the (unscaled) complementarity is less than
0317 %      this threshold.
0318 %
0319 % acceptable_tol                         0 <  (      1e-06) <  +inf
0320 %    "Acceptable" convergence tolerance (relative).
0321 %      Determines which (scaled) overall optimality error is considered to be
0322 %      "acceptable." There are two levels of termination criteria.  If the usual
0323 %      "desired" tolerances (see tol, dual_inf_tol etc) are satisfied at an
0324 %      iteration, the algorithm immediately terminates with a success message.
0325 %      On the other hand, if the algorithm encounters "acceptable_iter" many
0326 %      iterations in a row that are considered "acceptable", it will terminate
0327 %      before the desired convergence tolerance is met. This is useful in cases
0328 %      where the algorithm might not be able to achieve the "desired" level of
0329 %      accuracy.
0330 %
0331 % acceptable_iter                        0 <= (         15) <  +inf
0332 %    Number of "acceptable" iterates before triggering termination.
0333 %      If the algorithm encounters this many successive "acceptable" iterates
0334 %      (see "acceptable_tol"), it terminates, assuming that the problem has been
0335 %      solved to best possible accuracy given round-off.  If it is set to zero,
0336 %      this heuristic is disabled.
0337 %
0338 % acceptable_dual_inf_tol                0 <  (      1e+10) <  +inf
0339 %    "Acceptance" threshold for the dual infeasibility.
0340 %      Absolute tolerance on the dual infeasibility. "Acceptable" termination
0341 %      requires that the (max-norm of the unscaled) dual infeasibility is less
0342 %      than this threshold; see also acceptable_tol.
0343 %
0344 % acceptable_constr_viol_tol             0 <  (       0.01) <  +inf
0345 %    "Acceptance" threshold for the constraint violation.
0346 %      Absolute tolerance on the constraint violation. "Acceptable" termination
0347 %      requires that the max-norm of the (unscaled) constraint violation is less
0348 %      than this threshold; see also acceptable_tol.
0349 %
0350 % acceptable_compl_inf_tol               0 <  (       0.01) <  +inf
0351 %    "Acceptance" threshold for the complementarity conditions.
0352 %      Absolute tolerance on the complementarity. "Acceptable" termination
0353 %      requires that the max-norm of the (unscaled) complementarity is less than
0354 %      this threshold; see also acceptable_tol.
0355 %
0356 % acceptable_obj_change_tol              0 <= (      1e+20) <  +inf
0357 %    "Acceptance" stopping criterion based on objective function change.
0358 %      If the relative change of the objective function (scaled by
0359 %      Max(1,|f(x)|)) is less than this value, this part of the acceptable
0360 %      tolerance termination is satisfied; see also acceptable_tol.  This is
0361 %      useful for the quasi-Newton option, which has trouble to bring down the
0362 %      dual infeasibility.
0363 %
0364 % diverging_iterates_tol                 0 <  (      1e+20) <  +inf
0365 %    Threshold for maximal value of primal iterates.
0366 %      If any component of the primal iterates exceeded this value (in absolute
0367 %      terms), the optimization is aborted with the exit message that the
0368 %      iterates seem to be diverging.
0369 %
0370 %
0371 %
0372 % ### NLP Scaling ###
0373 %
0374 % nlp_scaling_method            ("gradient-based")
0375 %    Select the technique used for scaling the NLP.
0376 %      Selects the technique used for scaling the problem internally before it
0377 %      is solved. For user-scaling, the parameters come from the NLP. If you are
0378 %      using AMPL, they can be specified through suffixes ("scaling_factor")
0379 %    Possible values:
0380 %     - none                    [no problem scaling will be performed]
0381 %     - user-scaling            [scaling parameters will come from the user]
0382 %     - gradient-based          [scale the problem so the maximum gradient at
0383 %                                the starting point is scaling_max_gradient]
0384 %     - equilibration-based     [scale the problem so that first derivatives are
0385 %                                of order 1 at random points (only available
0386 %                                with MC19)]
0387 %
0388 % obj_scaling_factor                  -inf <  (          1) <  +inf
0389 %    Scaling factor for the objective function.
0390 %      This option sets a scaling factor for the objective function. The scaling
0391 %      is seen internally by Ipopt but the unscaled objective is reported in the
0392 %      console output. If additional scaling parameters are computed (e.g.
0393 %      user-scaling or gradient-based), both factors are multiplied. If this
0394 %      value is chosen to be negative, Ipopt will maximize the objective
0395 %      function instead of minimizing it.
0396 %
0397 % nlp_scaling_max_gradient               0 <  (        100) <  +inf
0398 %    Maximum gradient after NLP scaling.
0399 %      This is the gradient scaling cut-off. If the maximum gradient is above
0400 %      this value, then gradient based scaling will be performed. Scaling
0401 %      parameters are calculated to scale the maximum gradient back to this
0402 %      value. (This is g_max in Section 3.8 of the implementation paper.) Note:
0403 %      This option is only used if "nlp_scaling_method" is chosen as
0404 %      "gradient-based".
0405 %
0406 % nlp_scaling_obj_target_gradient         0 <= (          0) <  +inf
0407 %    Target value for objective function gradient size.
0408 %      If a positive number is chosen, the scaling factor the objective function
0409 %      is computed so that the gradient has the max norm of the given size at
0410 %      the starting point.  This overrides nlp_scaling_max_gradient for the
0411 %      objective function.
0412 %
0413 % nlp_scaling_constr_target_gradient         0 <= (          0) <  +inf
0414 %    Target value for constraint function gradient size.
0415 %      If a positive number is chosen, the scaling factor the constraint
0416 %      functions is computed so that the gradient has the max norm of the given
0417 %      size at the starting point.  This overrides nlp_scaling_max_gradient for
0418 %      the constraint functions.
0419 %
0420 %
0421 %
0422 % ### NLP ###
0423 %
0424 % nlp_lower_bound_inf                 -inf <  (     -1e+19) <  +inf
0425 %    any bound less or equal this value will be considered -inf (i.e. not lower
0426 %    bounded).
0427 %
0428 % nlp_upper_bound_inf                 -inf <  (      1e+19) <  +inf
0429 %    any bound greater or this value will be considered +inf (i.e. not upper
0430 %    bounded).
0431 %
0432 % fixed_variable_treatment      ("make_parameter")
0433 %    Determines how fixed variables should be handled.
0434 %      The main difference between those options is that the starting point in
0435 %      the "make_constraint" case still has the fixed variables at their given
0436 %      values, whereas in the case "make_parameter" the functions are always
0437 %      evaluated with the fixed values for those variables.  Also, for
0438 %      "relax_bounds", the fixing bound constraints are relaxed (according to"
0439 %      bound_relax_factor"). For both "make_constraints" and "relax_bounds",
0440 %      bound multipliers are computed for the fixed variables.
0441 %    Possible values:
0442 %     - make_parameter          [Remove fixed variable from optimization
0443 %                                variables]
0444 %     - make_constraint         [Add equality constraints fixing variables]
0445 %     - relax_bounds            [Relax fixing bound constraints]
0446 %
0447 % dependency_detector           ("none")
0448 %    Indicates which linear solver should be used to detect linearly dependent
0449 %    equality constraints.
0450 %      The default and available choices depend on how Ipopt has been compiled.
0451 %      This is experimental and does not work well.
0452 %    Possible values:
0453 %     - none                    [don't check; no extra work at beginning]
0454 %     - mumps                   [use MUMPS]
0455 %     - wsmp                    [use WSMP]
0456 %     - ma28                    [use MA28]
0457 %
0458 % dependency_detection_with_rhs ("no")
0459 %    Indicates if the right hand sides of the constraints should be considered
0460 %    during dependency detection
0461 %    Possible values:
0462 %     - no                      [only look at gradients]
0463 %     - yes                     [also consider right hand side]
0464 %
0465 % num_linear_variables                   0 <= (          0) <  +inf
0466 %    Number of linear variables
0467 %      When the Hessian is approximated, it is assumed that the first
0468 %      num_linear_variables variables are linear.  The Hessian is then not
0469 %      approximated in this space.  If the get_number_of_nonlinear_variables
0470 %      method in the TNLP is implemented, this option is ignored.
0471 %
0472 % kappa_d                                0 <= (      1e-05) <  +inf
0473 %    Weight for linear damping term (to handle one-sided bounds).
0474 %      (see Section 3.7 in implementation paper.)
0475 %
0476 % bound_relax_factor                     0 <= (      1e-08) <  +inf
0477 %    Factor for initial relaxation of the bounds.
0478 %      Before start of the optimization, the bounds given by the user are
0479 %      relaxed.  This option sets the factor for this relaxation.  If it is set
0480 %      to zero, then then bounds relaxation is disabled. (See Eqn.(35) in
0481 %      implementation paper.)
0482 %
0483 % honor_original_bounds         ("yes")
0484 %    Indicates whether final points should be projected into original bounds.
0485 %      Ipopt might relax the bounds during the optimization (see, e.g., option
0486 %      "bound_relax_factor").  This option determines whether the final point
0487 %      should be projected back into the user-provide original bounds after the
0488 %      optimization.
0489 %    Possible values:
0490 %     - no                      [Leave final point unchanged]
0491 %     - yes                     [Project final point back into original bounds]
0492 %
0493 % check_derivatives_for_naninf  ("no")
0494 %    Indicates whether it is desired to check for Nan/Inf in derivative matrices
0495 %      Activating this option will cause an error if an invalid number is
0496 %      detected in the constraint Jacobians or the Lagrangian Hessian.  If this
0497 %      is not activated, the test is skipped, and the algorithm might proceed
0498 %      with invalid numbers and fail.
0499 %    Possible values:
0500 %     - no                      [Don't check (faster).]
0501 %     - yes                     [Check Jacobians and Hessian for Nan and Inf.]
0502 %
0503 % jac_c_constant                ("no")
0504 %    Indicates whether all equality constraints are linear
0505 %      Activating this option will cause Ipopt to ask for the Jacobian of the
0506 %      equality constraints only once from the NLP and reuse this information
0507 %      later.
0508 %    Possible values:
0509 %     - no                      [Don't assume that all equality constraints are
0510 %                                linear]
0511 %     - yes                     [Assume that equality constraints Jacobian are
0512 %                                constant]
0513 %
0514 % jac_d_constant                ("no")
0515 %    Indicates whether all inequality constraints are linear
0516 %      Activating this option will cause Ipopt to ask for the Jacobian of the
0517 %      inequality constraints only once from the NLP and reuse this information
0518 %      later.
0519 %    Possible values:
0520 %     - no                      [Don't assume that all inequality constraints
0521 %                                are linear]
0522 %     - yes                     [Assume that equality constraints Jacobian are
0523 %                                constant]
0524 %
0525 % hessian_constant              ("no")
0526 %    Indicates whether the problem is a quadratic problem
0527 %      Activating this option will cause Ipopt to ask for the Hessian of the
0528 %      Lagrangian function only once from the NLP and reuse this information
0529 %      later.
0530 %    Possible values:
0531 %     - no                      [Assume that Hessian changes]
0532 %     - yes                     [Assume that Hessian is constant]
0533 %
0534 %
0535 %
0536 % ### Initialization ###
0537 %
0538 % bound_push                             0 <  (       0.01) <  +inf
0539 %    Desired minimum absolute distance from the initial point to bound.
0540 %      Determines how much the initial point might have to be modified in order
0541 %      to be sufficiently inside the bounds (together with "bound_frac").  (This
0542 %      is kappa_1 in Section 3.6 of implementation paper.)
0543 %
0544 % bound_frac                             0 <  (       0.01) <= 0.5
0545 %    Desired minimum relative distance from the initial point to bound.
0546 %      Determines how much the initial point might have to be modified in order
0547 %      to be sufficiently inside the bounds (together with "bound_push").  (This
0548 %      is kappa_2 in Section 3.6 of implementation paper.)
0549 %
0550 % slack_bound_push                       0 <  (       0.01) <  +inf
0551 %    Desired minimum absolute distance from the initial slack to bound.
0552 %      Determines how much the initial slack variables might have to be modified
0553 %      in order to be sufficiently inside the inequality bounds (together with
0554 %      "slack_bound_frac").  (This is kappa_1 in Section 3.6 of implementation
0555 %      paper.)
0556 %
0557 % slack_bound_frac                       0 <  (       0.01) <= 0.5
0558 %    Desired minimum relative distance from the initial slack to bound.
0559 %      Determines how much the initial slack variables might have to be modified
0560 %      in order to be sufficiently inside the inequality bounds (together with
0561 %      "slack_bound_push").  (This is kappa_2 in Section 3.6 of implementation
0562 %      paper.)
0563 %
0564 % constr_mult_init_max                   0 <= (       1000) <  +inf
0565 %    Maximum allowed least-square guess of constraint multipliers.
0566 %      Determines how large the initial least-square guesses of the constraint
0567 %      multipliers are allowed to be (in max-norm). If the guess is larger than
0568 %      this value, it is discarded and all constraint multipliers are set to
0569 %      zero.  This options is also used when initializing the restoration phase.
0570 %      By default, "resto.constr_mult_init_max" (the one used in
0571 %      RestoIterateInitializer) is set to zero.
0572 %
0573 % bound_mult_init_val                    0 <  (          1) <  +inf
0574 %    Initial value for the bound multipliers.
0575 %      All dual variables corresponding to bound constraints are initialized to
0576 %      this value.
0577 %
0578 % bound_mult_init_method        ("constant")
0579 %    Initialization method for bound multipliers
0580 %      This option defines how the iterates for the bound multipliers are
0581 %      initialized.  If "constant" is chosen, then all bound multipliers are
0582 %      initialized to the value of "bound_mult_init_val".  If "mu-based" is
0583 %      chosen, the each value is initialized to the the value of "mu_init"
0584 %      divided by the corresponding slack variable.  This latter option might be
0585 %      useful if the starting point is close to the optimal solution.
0586 %    Possible values:
0587 %     - constant                [set all bound multipliers to the value of
0588 %                                bound_mult_init_val]
0589 %     - mu-based                [initialize to mu_init/x_slack]
0590 %
0591 % least_square_init_primal      ("no")
0592 %    Least square initialization of the primal variables
0593 %      If set to yes, Ipopt ignores the user provided point and solves a least
0594 %      square problem for the primal variables (x and s), to fit the linearized
0595 %      equality and inequality constraints.  This might be useful if the user
0596 %      doesn't know anything about the starting point, or for solving an LP or
0597 %      QP.
0598 %    Possible values:
0599 %     - no                      [take user-provided point]
0600 %     - yes                     [overwrite user-provided point with least-square
0601 %                                estimates]
0602 %
0603 % least_square_init_duals       ("no")
0604 %    Least square initialization of all dual variables
0605 %      If set to yes, Ipopt tries to compute least-square multipliers
0606 %      (considering ALL dual variables).  If successful, the bound multipliers
0607 %      are possibly corrected to be at least bound_mult_init_val. This might be
0608 %      useful if the user doesn't know anything about the starting point, or for
0609 %      solving an LP or QP.  This overwrites option "bound_mult_init_method".
0610 %    Possible values:
0611 %     - no                      [use bound_mult_init_val and least-square
0612 %                                equality constraint multipliers]
0613 %     - yes                     [overwrite user-provided point with least-square
0614 %                                estimates]
0615 %
0616 %
0617 %
0618 % ### Barrier Parameter Update ###
0619 %
0620 % mu_max_fact                            0 <  (       1000) <  +inf
0621 %    Factor for initialization of maximum value for barrier parameter.
0622 %      This option determines the upper bound on the barrier parameter.  This
0623 %      upper bound is computed as the average complementarity at the initial
0624 %      point times the value of this option. (Only used if option "mu_strategy"
0625 %      is chosen as "adaptive".)
0626 %
0627 % mu_max                                 0 <  (     100000) <  +inf
0628 %    Maximum value for barrier parameter.
0629 %      This option specifies an upper bound on the barrier parameter in the
0630 %      adaptive mu selection mode.  If this option is set, it overwrites the
0631 %      effect of mu_max_fact. (Only used if option "mu_strategy" is chosen as
0632 %      "adaptive".)
0633 %
0634 % mu_min                                 0 <  (      1e-11) <  +inf
0635 %    Minimum value for barrier parameter.
0636 %      This option specifies the lower bound on the barrier parameter in the
0637 %      adaptive mu selection mode. By default, it is set to the minimum of 1e-11
0638 %      and min("tol","compl_inf_tol")/("barrier_tol_factor"+1), which should be
0639 %      a reasonable value. (Only used if option "mu_strategy" is chosen as
0640 %      "adaptive".)
0641 %
0642 % adaptive_mu_globalization     ("obj-constr-filter")
0643 %    Globalization strategy for the adaptive mu selection mode.
0644 %      To achieve global convergence of the adaptive version, the algorithm has
0645 %      to switch to the monotone mode (Fiacco-McCormick approach) when
0646 %      convergence does not seem to appear.  This option sets the criterion used
0647 %      to decide when to do this switch. (Only used if option "mu_strategy" is
0648 %      chosen as "adaptive".)
0649 %    Possible values:
0650 %     - kkt-error               [nonmonotone decrease of kkt-error]
0651 %     - obj-constr-filter       [2-dim filter for objective and constraint
0652 %                                violation]
0653 %     - never-monotone-mode     [disables globalization]
0654 %
0655 % adaptive_mu_kkterror_red_iters         0 <= (          4) <  +inf
0656 %    Maximum number of iterations requiring sufficient progress.
0657 %      For the "kkt-error" based globalization strategy, sufficient progress
0658 %      must be made for "adaptive_mu_kkterror_red_iters" iterations. If this
0659 %      number of iterations is exceeded, the globalization strategy switches to
0660 %      the monotone mode.
0661 %
0662 % adaptive_mu_kkterror_red_fact          0 <  (     0.9999) <  1
0663 %    Sufficient decrease factor for "kkt-error" globalization strategy.
0664 %      For the "kkt-error" based globalization strategy, the error must decrease
0665 %      by this factor to be deemed sufficient decrease.
0666 %
0667 % filter_margin_fact                     0 <  (      1e-05) <  1
0668 %    Factor determining width of margin for obj-constr-filter adaptive
0669 %    globalization strategy.
0670 %      When using the adaptive globalization strategy, "obj-constr-filter",
0671 %      sufficient progress for a filter entry is defined as follows: (new obj) <
0672 %      (filter obj) - filter_margin_fact*(new constr-viol) OR (new constr-viol)
0673 %      < (filter constr-viol) - filter_margin_fact*(new constr-viol).  For the
0674 %      description of the "kkt-error-filter" option see "filter_max_margin".
0675 %
0676 % filter_max_margin                      0 <  (          1) <  +inf
0677 %    Maximum width of margin in obj-constr-filter adaptive globalization
0678 %    strategy.
0679 %
0680 % adaptive_mu_restore_previous_iterate("no")
0681 %    Indicates if the previous iterate should be restored if the monotone mode
0682 %    is entered.
0683 %      When the globalization strategy for the adaptive barrier algorithm
0684 %      switches to the monotone mode, it can either start from the most recent
0685 %      iterate (no), or from the last iterate that was accepted (yes).
0686 %    Possible values:
0687 %     - no                      [don't restore accepted iterate]
0688 %     - yes                     [restore accepted iterate]
0689 %
0690 % adaptive_mu_monotone_init_factor         0 <  (        0.8) <  +inf
0691 %    Determines the initial value of the barrier parameter when switching to the
0692 %    monotone mode.
0693 %      When the globalization strategy for the adaptive barrier algorithm
0694 %      switches to the monotone mode and fixed_mu_oracle is chosen as
0695 %      "average_compl", the barrier parameter is set to the current average
0696 %      complementarity times the value of "adaptive_mu_monotone_init_factor".
0697 %
0698 % adaptive_mu_kkt_norm_type     ("2-norm-squared")
0699 %    Norm used for the KKT error in the adaptive mu globalization strategies.
0700 %      When computing the KKT error for the globalization strategies, the norm
0701 %      to be used is specified with this option. Note, this options is also used
0702 %      in the QualityFunctionMuOracle.
0703 %    Possible values:
0704 %     - 1-norm                  [use the 1-norm (abs sum)]
0705 %     - 2-norm-squared          [use the 2-norm squared (sum of squares)]
0706 %     - max-norm                [use the infinity norm (max)]
0707 %     - 2-norm                  [use 2-norm]
0708 %
0709 % mu_strategy                   ("monotone")
0710 %    Update strategy for barrier parameter.
0711 %      Determines which barrier parameter update strategy is to be used.
0712 %    Possible values:
0713 %     - monotone                [use the monotone (Fiacco-McCormick) strategy]
0714 %     - adaptive                [use the adaptive update strategy]
0715 %
0716 % mu_oracle                     ("quality-function")
0717 %    Oracle for a new barrier parameter in the adaptive strategy.
0718 %      Determines how a new barrier parameter is computed in each "free-mode"
0719 %      iteration of the adaptive barrier parameter strategy. (Only considered if
0720 %      "adaptive" is selected for option "mu_strategy").
0721 %    Possible values:
0722 %     - probing                 [Mehrotra's probing heuristic]
0723 %     - loqo                    [LOQO's centrality rule]
0724 %     - quality-function        [minimize a quality function]
0725 %
0726 % fixed_mu_oracle               ("average_compl")
0727 %    Oracle for the barrier parameter when switching to fixed mode.
0728 %      Determines how the first value of the barrier parameter should be
0729 %      computed when switching to the "monotone mode" in the adaptive strategy.
0730 %      (Only considered if "adaptive" is selected for option "mu_strategy".)
0731 %    Possible values:
0732 %     - probing                 [Mehrotra's probing heuristic]
0733 %     - loqo                    [LOQO's centrality rule]
0734 %     - quality-function        [minimize a quality function]
0735 %     - average_compl           [base on current average complementarity]
0736 %
0737 % mu_init                                0 <  (        0.1) <  +inf
0738 %    Initial value for the barrier parameter.
0739 %      This option determines the initial value for the barrier parameter (mu).
0740 %      It is only relevant in the monotone, Fiacco-McCormick version of the
0741 %      algorithm. (i.e., if "mu_strategy" is chosen as "monotone")
0742 %
0743 % barrier_tol_factor                     0 <  (         10) <  +inf
0744 %    Factor for mu in barrier stop test.
0745 %      The convergence tolerance for each barrier problem in the monotone mode
0746 %      is the value of the barrier parameter times "barrier_tol_factor". This
0747 %      option is also used in the adaptive mu strategy during the monotone mode.
0748 %      (This is kappa_epsilon in implementation paper).
0749 %
0750 % mu_linear_decrease_factor              0 <  (        0.2) <  1
0751 %    Determines linear decrease rate of barrier parameter.
0752 %      For the Fiacco-McCormick update procedure the new barrier parameter mu is
0753 %      obtained by taking the minimum of mu*"mu_linear_decrease_factor" and
0754 %      mu^"superlinear_decrease_power".  (This is kappa_mu in implementation
0755 %      paper.) This option is also used in the adaptive mu strategy during the
0756 %      monotone mode.
0757 %
0758 % mu_superlinear_decrease_power          1 <  (        1.5) <  2
0759 %    Determines superlinear decrease rate of barrier parameter.
0760 %      For the Fiacco-McCormick update procedure the new barrier parameter mu is
0761 %      obtained by taking the minimum of mu*"mu_linear_decrease_factor" and
0762 %      mu^"superlinear_decrease_power".  (This is theta_mu in implementation
0763 %      paper.) This option is also used in the adaptive mu strategy during the
0764 %      monotone mode.
0765 %
0766 % mu_allow_fast_monotone_decrease("yes")
0767 %    Allow skipping of barrier problem if barrier test is already met.
0768 %      If set to "no", the algorithm enforces at least one iteration per barrier
0769 %      problem, even if the barrier test is already met for the updated barrier
0770 %      parameter.
0771 %    Possible values:
0772 %     - no                      [Take at least one iteration per barrier problem]
0773 %     - yes                     [Allow fast decrease of mu if barrier test it met]
0774 %
0775 % tau_min                                0 <  (       0.99) <  1
0776 %    Lower bound on fraction-to-the-boundary parameter tau.
0777 %      (This is tau_min in the implementation paper.)  This option is also used
0778 %      in the adaptive mu strategy during the monotone mode.
0779 %
0780 % sigma_max                              0 <  (        100) <  +inf
0781 %    Maximum value of the centering parameter.
0782 %      This is the upper bound for the centering parameter chosen by the quality
0783 %      function based barrier parameter update. (Only used if option "mu_oracle"
0784 %      is set to "quality-function".)
0785 %
0786 % sigma_min                              0 <= (      1e-06) <  +inf
0787 %    Minimum value of the centering parameter.
0788 %      This is the lower bound for the centering parameter chosen by the quality
0789 %      function based barrier parameter update. (Only used if option "mu_oracle"
0790 %      is set to "quality-function".)
0791 %
0792 % quality_function_norm_type    ("2-norm-squared")
0793 %    Norm used for components of the quality function.
0794 %      (Only used if option "mu_oracle" is set to "quality-function".)
0795 %    Possible values:
0796 %     - 1-norm                  [use the 1-norm (abs sum)]
0797 %     - 2-norm-squared          [use the 2-norm squared (sum of squares)]
0798 %     - max-norm                [use the infinity norm (max)]
0799 %     - 2-norm                  [use 2-norm]
0800 %
0801 % quality_function_centrality   ("none")
0802 %    The penalty term for centrality that is included in quality function.
0803 %      This determines whether a term is added to the quality function to
0804 %      penalize deviation from centrality with respect to complementarity.  The
0805 %      complementarity measure here is the xi in the Loqo update rule. (Only
0806 %      used if option "mu_oracle" is set to "quality-function".)
0807 %    Possible values:
0808 %     - none                    [no penalty term is added]
0809 %     - log                     [complementarity * the log of the centrality
0810 %                                measure]
0811 %     - reciprocal              [complementarity * the reciprocal of the
0812 %                                centrality measure]
0813 %     - cubed-reciprocal        [complementarity * the reciprocal of the
0814 %                                centrality measure cubed]
0815 %
0816 % quality_function_balancing_term("none")
0817 %    The balancing term included in the quality function for centrality.
0818 %      This determines whether a term is added to the quality function that
0819 %      penalizes situations where the complementarity is much smaller than dual
0820 %      and primal infeasibilities. (Only used if option "mu_oracle" is set to
0821 %      "quality-function".)
0822 %    Possible values:
0823 %     - none                    [no balancing term is added]
0824 %     - cubic                   [Max(0,Max(dual_inf,primal_inf)-compl)^3]
0825 %
0826 % quality_function_max_section_steps         0 <= (          8) <  +inf
0827 %    Maximum number of search steps during direct search procedure determining
0828 %    the optimal centering parameter.
0829 %      The golden section search is performed for the quality function based mu
0830 %      oracle. (Only used if option "mu_oracle" is set to "quality-function".)
0831 %
0832 % quality_function_section_sigma_tol         0 <= (       0.01) <  1
0833 %    Tolerance for the section search procedure determining the optimal
0834 %    centering parameter (in sigma space).
0835 %      The golden section search is performed for the quality function based mu
0836 %      oracle. (Only used if option "mu_oracle" is set to "quality-function".)
0837 %
0838 % quality_function_section_qf_tol         0 <= (          0) <  1
0839 %    Tolerance for the golden section search procedure determining the optimal
0840 %    centering parameter (in the function value space).
0841 %      The golden section search is performed for the quality function based mu
0842 %      oracle. (Only used if option "mu_oracle" is set to "quality-function".)
0843 %
0844 %
0845 %
0846 % ### Line Search ###
0847 %
0848 % alpha_red_factor                       0 <  (        0.5) <  1
0849 %    Fractional reduction of the trial step size in the backtracking line search.
0850 %      At every step of the backtracking line search, the trial step size is
0851 %      reduced by this factor.
0852 %
0853 % accept_every_trial_step       ("no")
0854 %    Always accept the first trial step.
0855 %      Setting this option to "yes" essentially disables the line search and
0856 %      makes the algorithm take aggressive steps, without global convergence
0857 %      guarantees.
0858 %    Possible values:
0859 %     - no                      [don't arbitrarily accept the full step]
0860 %     - yes                     [always accept the full step]
0861 %
0862 % accept_after_max_steps                -1 <= (         -1) <  +inf
0863 %    Accept a trial point after maximal this number of steps.
0864 %      Even if it does not satisfy line search conditions.
0865 %
0866 % alpha_for_y                   ("primal")
0867 %    Method to determine the step size for constraint multipliers.
0868 %      This option determines how the step size (alpha_y) will be calculated
0869 %      when updating the constraint multipliers.
0870 %    Possible values:
0871 %     - primal                  [use primal step size]
0872 %     - bound-mult              [use step size for the bound multipliers (good
0873 %                                for LPs)]
0874 %     - min                     [use the min of primal and bound multipliers]
0875 %     - max                     [use the max of primal and bound multipliers]
0876 %     - full                    [take a full step of size one]
0877 %     - min-dual-infeas         [choose step size minimizing new dual
0878 %                                infeasibility]
0879 %     - safer-min-dual-infeas   [like "min_dual_infeas", but safeguarded by
0880 %                                "min" and "max"]
0881 %     - primal-and-full         [use the primal step size, and full step if
0882 %                                delta_x <= alpha_for_y_tol]
0883 %     - dual-and-full           [use the dual step size, and full step if
0884 %                                delta_x <= alpha_for_y_tol]
0885 %     - acceptor                [Call LSAcceptor to get step size for y]
0886 %
0887 % alpha_for_y_tol                        0 <= (         10) <  +inf
0888 %    Tolerance for switching to full equality multiplier steps.
0889 %      This is only relevant if "alpha_for_y" is chosen "primal-and-full" or
0890 %      "dual-and-full".  The step size for the equality constraint multipliers
0891 %      is taken to be one if the max-norm of the primal step is less than this
0892 %      tolerance.
0893 %
0894 % tiny_step_tol                          0 <= (2.22045e-15) <  +inf
0895 %    Tolerance for detecting numerically insignificant steps.
0896 %      If the search direction in the primal variables (x and s) is, in relative
0897 %      terms for each component, less than this value, the algorithm accepts the
0898 %      full step without line search.  If this happens repeatedly, the algorithm
0899 %      will terminate with a corresponding exit message. The default value is 10
0900 %      times machine precision.
0901 %
0902 % tiny_step_y_tol                        0 <= (       0.01) <  +inf
0903 %    Tolerance for quitting because of numerically insignificant steps.
0904 %      If the search direction in the primal variables (x and s) is, in relative
0905 %      terms for each component, repeatedly less than tiny_step_tol, and the
0906 %      step in the y variables is smaller than this threshold, the algorithm
0907 %      will terminate.
0908 %
0909 % watchdog_shortened_iter_trigger         0 <= (         10) <  +inf
0910 %    Number of shortened iterations that trigger the watchdog.
0911 %      If the number of successive iterations in which the backtracking line
0912 %      search did not accept the first trial point exceeds this number, the
0913 %      watchdog procedure is activated.  Choosing "0" here disables the watchdog
0914 %      procedure.
0915 %
0916 % watchdog_trial_iter_max                1 <= (          3) <  +inf
0917 %    Maximum number of watchdog iterations.
0918 %      This option determines the number of trial iterations allowed before the
0919 %      watchdog procedure is aborted and the algorithm returns to the stored
0920 %      point.
0921 %
0922 % theta_max_fact                         0 <  (      10000) <  +inf
0923 %    Determines upper bound for constraint violation in the filter.
0924 %      The algorithmic parameter theta_max is determined as theta_max_fact times
0925 %      the maximum of 1 and the constraint violation at initial point.  Any
0926 %      point with a constraint violation larger than theta_max is unacceptable
0927 %      to the filter (see Eqn. (21) in the implementation paper).
0928 %
0929 % theta_min_fact                         0 <  (     0.0001) <  +inf
0930 %    Determines constraint violation threshold in the switching rule.
0931 %      The algorithmic parameter theta_min is determined as theta_min_fact times
0932 %      the maximum of 1 and the constraint violation at initial point.  The
0933 %      switching rules treats an iteration as an h-type iteration whenever the
0934 %      current constraint violation is larger than theta_min (see paragraph
0935 %      before Eqn. (19) in the implementation paper).
0936 %
0937 % eta_phi                                0 <  (      1e-08) <  0.5
0938 %    Relaxation factor in the Armijo condition.
0939 %      (See Eqn. (20) in the implementation paper)
0940 %
0941 % delta                                  0 <  (          1) <  +inf
0942 %    Multiplier for constraint violation in the switching rule.
0943 %      (See Eqn. (19) in the implementation paper.)
0944 %
0945 % s_phi                                  1 <  (        2.3) <  +inf
0946 %    Exponent for linear barrier function model in the switching rule.
0947 %      (See Eqn. (19) in the implementation paper.)
0948 %
0949 % s_theta                                1 <  (        1.1) <  +inf
0950 %    Exponent for current constraint violation in the switching rule.
0951 %      (See Eqn. (19) in the implementation paper.)
0952 %
0953 % gamma_phi                              0 <  (      1e-08) <  1
0954 %    Relaxation factor in the filter margin for the barrier function.
0955 %      (See Eqn. (18a) in the implementation paper.)
0956 %
0957 % gamma_theta                            0 <  (      1e-05) <  1
0958 %    Relaxation factor in the filter margin for the constraint violation.
0959 %      (See Eqn. (18b) in the implementation paper.)
0960 %
0961 % alpha_min_frac                         0 <  (       0.05) <  1
0962 %    Safety factor for the minimal step size (before switching to restoration
0963 %    phase).
0964 %      (This is gamma_alpha in Eqn. (20) in the implementation paper.)
0965 %
0966 % max_soc                                0 <= (          4) <  +inf
0967 %    Maximum number of second order correction trial steps at each iteration.
0968 %      Choosing 0 disables the second order corrections. (This is p^{max} of
0969 %      Step A-5.9 of Algorithm A in the implementation paper.)
0970 %
0971 % kappa_soc                              0 <  (       0.99) <  +inf
0972 %    Factor in the sufficient reduction rule for second order correction.
0973 %      This option determines how much a second order correction step must
0974 %      reduce the constraint violation so that further correction steps are
0975 %      attempted.  (See Step A-5.9 of Algorithm A in the implementation paper.)
0976 %
0977 % obj_max_inc                            1 <  (          5) <  +inf
0978 %    Determines the upper bound on the acceptable increase of barrier objective
0979 %    function.
0980 %      Trial points are rejected if they lead to an increase in the barrier
0981 %      objective function by more than obj_max_inc orders of magnitude.
0982 %
0983 % max_filter_resets                      0 <= (          5) <  +inf
0984 %    Maximal allowed number of filter resets
0985 %      A positive number enables a heuristic that resets the filter, whenever in
0986 %      more than "filter_reset_trigger" successive iterations the last rejected
0987 %      trial steps size was rejected because of the filter.  This option
0988 %      determine the maximal number of resets that are allowed to take place.
0989 %
0990 % filter_reset_trigger                   1 <= (          5) <  +inf
0991 %    Number of iterations that trigger the filter reset.
0992 %      If the filter reset heuristic is active and the number of successive
0993 %      iterations in which the last rejected trial step size was rejected
0994 %      because of the filter, the filter is reset.
0995 %
0996 % corrector_type                ("none")
0997 %    The type of corrector steps that should be taken (unsupported!).
0998 %      If "mu_strategy" is "adaptive", this option determines what kind of
0999 %      corrector steps should be tried.
1000 %    Possible values:
1001 %     - none                    [no corrector]
1002 %     - affine                  [corrector step towards mu=0]
1003 %     - primal-dual             [corrector step towards current mu]
1004 %
1005 % skip_corr_if_neg_curv         ("yes")
1006 %    Skip the corrector step in negative curvature iteration (unsupported!).
1007 %      The corrector step is not tried if negative curvature has been
1008 %      encountered during the computation of the search direction in the current
1009 %      iteration. This option is only used if "mu_strategy" is "adaptive".
1010 %    Possible values:
1011 %     - no                      [don't skip]
1012 %     - yes                     [skip]
1013 %
1014 % skip_corr_in_monotone_mode    ("yes")
1015 %    Skip the corrector step during monotone barrier parameter mode
1016 %    (unsupported!).
1017 %      The corrector step is not tried if the algorithm is currently in the
1018 %      monotone mode (see also option "barrier_strategy").This option is only
1019 %      used if "mu_strategy" is "adaptive".
1020 %    Possible values:
1021 %     - no                      [don't skip]
1022 %     - yes                     [skip]
1023 %
1024 % corrector_compl_avrg_red_fact          0 <  (          1) <  +inf
1025 %    Complementarity tolerance factor for accepting corrector step
1026 %    (unsupported!).
1027 %      This option determines the factor by which complementarity is allowed to
1028 %      increase for a corrector step to be accepted.
1029 %
1030 % nu_init                                0 <  (      1e-06) <  +inf
1031 %    Initial value of the penalty parameter.
1032 %
1033 % nu_inc                                 0 <  (     0.0001) <  +inf
1034 %    Increment of the penalty parameter.
1035 %
1036 % rho                                    0 <  (        0.1) <  1
1037 %    Value in penalty parameter update formula.
1038 %
1039 % kappa_sigma                            0 <  (      1e+10) <  +inf
1040 %    Factor limiting the deviation of dual variables from primal estimates.
1041 %      If the dual variables deviate from their primal estimates, a correction
1042 %      is performed. (See Eqn. (16) in the implementation paper.) Setting the
1043 %      value to less than 1 disables the correction.
1044 %
1045 % recalc_y                      ("no")
1046 %    Tells the algorithm to recalculate the equality and inequality multipliers
1047 %    as least square estimates.
1048 %      This asks the algorithm to recompute the multipliers, whenever the
1049 %      current infeasibility is less than recalc_y_feas_tol. Choosing yes might
1050 %      be helpful in the quasi-Newton option.  However, each recalculation
1051 %      requires an extra factorization of the linear system.  If a limited
1052 %      memory quasi-Newton option is chosen, this is used by default.
1053 %    Possible values:
1054 %     - no                      [use the Newton step to update the multipliers]
1055 %     - yes                     [use least-square multiplier estimates]
1056 %
1057 % recalc_y_feas_tol                      0 <  (      1e-06) <  +inf
1058 %    Feasibility threshold for recomputation of multipliers.
1059 %      If recalc_y is chosen and the current infeasibility is less than this
1060 %      value, then the multipliers are recomputed.
1061 %
1062 % slack_move                             0 <= (1.81899e-12) <  +inf
1063 %    Correction size for very small slacks.
1064 %      Due to numerical issues or the lack of an interior, the slack variables
1065 %      might become very small.  If a slack becomes very small compared to
1066 %      machine precision, the corresponding bound is moved slightly.  This
1067 %      parameter determines how large the move should be.  Its default value is
1068 %      mach_eps^{3/4}.  (See also end of Section 3.5 in implementation paper -
1069 %      but actual implementation might be somewhat different.)
1070 %
1071 %
1072 %
1073 % ### Warm Start ###
1074 %
1075 % warm_start_init_point         ("no")
1076 %    Warm-start for initial point
1077 %      Indicates whether this optimization should use a warm start
1078 %      initialization, where values of primal and dual variables are given
1079 %      (e.g., from a previous optimization of a related problem.)
1080 %    Possible values:
1081 %     - no                      [do not use the warm start initialization]
1082 %     - yes                     [use the warm start initialization]
1083 %
1084 % warm_start_same_structure     ("no")
1085 %    Indicates whether a problem with a structure identical to the previous one
1086 %    is to be solved.
1087 %      If "yes" is chosen, then the algorithm assumes that an NLP is now to be
1088 %      solved, whose structure is identical to one that already was considered
1089 %      (with the same NLP object).
1090 %    Possible values:
1091 %     - no                      [Assume this is a new problem.]
1092 %     - yes                     [Assume this is problem has known structure]
1093 %
1094 % warm_start_bound_push                  0 <  (      0.001) <  +inf
1095 %    same as bound_push for the regular initializer.
1096 %
1097 % warm_start_bound_frac                  0 <  (      0.001) <= 0.5
1098 %    same as bound_frac for the regular initializer.
1099 %
1100 % warm_start_slack_bound_push            0 <  (      0.001) <  +inf
1101 %    same as slack_bound_push for the regular initializer.
1102 %
1103 % warm_start_slack_bound_frac            0 <  (      0.001) <= 0.5
1104 %    same as slack_bound_frac for the regular initializer.
1105 %
1106 % warm_start_mult_bound_push             0 <  (      0.001) <  +inf
1107 %    same as mult_bound_push for the regular initializer.
1108 %
1109 % warm_start_mult_init_max            -inf <  (      1e+06) <  +inf
1110 %    Maximum initial value for the equality multipliers.
1111 %
1112 % warm_start_entire_iterate     ("no")
1113 %    Tells algorithm whether to use the GetWarmStartIterate method in the NLP.
1114 %    Possible values:
1115 %     - no                      [call GetStartingPoint in the NLP]
1116 %     - yes                     [call GetWarmStartIterate in the NLP]
1117 %
1118 %
1119 %
1120 % ### Linear Solver ###
1121 %
1122 % linear_solver                 ("mumps")
1123 %    Linear solver used for step computations.
1124 %      Determines which linear algebra package is to be used for the solution of
1125 %      the augmented linear system (for obtaining the search directions). Note,
1126 %      the code must have been compiled with the linear solver you want to
1127 %      choose. Depending on your Ipopt installation, not all options are
1128 %      available.
1129 %    Possible values:
1130 %     - ma27                    [use the Harwell routine MA27]
1131 %     - ma57                    [use the Harwell routine MA57]
1132 %     - pardiso                 [use the Pardiso package]
1133 %     - wsmp                    [use WSMP package]
1134 %     - mumps                   [use MUMPS package]
1135 %     - custom                  [use custom linear solver]
1136 %
1137 % linear_system_scaling         ("none")
1138 %    Method for scaling the linear system.
1139 %      Determines the method used to compute symmetric scaling factors for the
1140 %      augmented system (see also the "linear_scaling_on_demand" option).  This
1141 %      scaling is independent of the NLP problem scaling.  By default, MC19 is
1142 %      only used if MA27 or MA57 are selected as linear solvers. This option is
1143 %      only available if Ipopt has been compiled with MC19.
1144 %    Possible values:
1145 %     - none                    [no scaling will be performed]
1146 %     - mc19                    [use the Harwell routine MC19]
1147 %
1148 % linear_scaling_on_demand      ("yes")
1149 %    Flag indicating that linear scaling is only done if it seems required.
1150 %      This option is only important if a linear scaling method (e.g., mc19) is
1151 %      used.  If you choose "no", then the scaling factors are computed for
1152 %      every linear system from the start.  This can be quite expensive.
1153 %      Choosing "yes" means that the algorithm will start the scaling method
1154 %      only when the solutions to the linear system seem not good, and then use
1155 %      it until the end.
1156 %    Possible values:
1157 %     - no                      [Always scale the linear system.]
1158 %     - yes                     [Start using linear system scaling if solutions
1159 %                                seem not good.]
1160 %
1161 %
1162 %
1163 % ### Step Calculation ###
1164 %
1165 % mehrotra_algorithm            ("no")
1166 %    Indicates if we want to do Mehrotra's algorithm.
1167 %      If set to yes, Ipopt runs as Mehrotra's predictor-corrector algorithm.
1168 %      This works usually very well for LPs and convex QPs.  This automatically
1169 %      disables the line search, and chooses the (unglobalized) adaptive mu
1170 %      strategy with the "probing" oracle, and uses "corrector_type=affine"
1171 %      without any safeguards; you should not set any of those options
1172 %      explicitly in addition.  Also, unless otherwise specified, the values of
1173 %      "bound_push", "bound_frac", and "bound_mult_init_val" are set more
1174 %      aggressive, and sets "alpha_for_y=bound_mult".
1175 %    Possible values:
1176 %     - no                      [Do the usual Ipopt algorithm.]
1177 %     - yes                     [Do Mehrotra's predictor-corrector algorithm.]
1178 %
1179 % fast_step_computation         ("no")
1180 %    Indicates if the linear system should be solved quickly.
1181 %      If set to yes, the algorithm assumes that the linear system that is
1182 %      solved to obtain the search direction, is solved sufficiently well. In
1183 %      that case, no residuals are computed, and the computation of the search
1184 %      direction is a little faster.
1185 %    Possible values:
1186 %     - no                      [Verify solution of linear system by computing
1187 %                                residuals.]
1188 %     - yes                     [Trust that linear systems are solved well.]
1189 %
1190 % min_refinement_steps                   0 <= (          1) <  +inf
1191 %    Minimum number of iterative refinement steps per linear system solve.
1192 %      Iterative refinement (on the full unsymmetric system) is performed for
1193 %      each right hand side.  This option determines the minimum number of
1194 %      iterative refinements (i.e. at least "min_refinement_steps" iterative
1195 %      refinement steps are enforced per right hand side.)
1196 %
1197 % max_refinement_steps                   0 <= (         10) <  +inf
1198 %    Maximum number of iterative refinement steps per linear system solve.
1199 %      Iterative refinement (on the full unsymmetric system) is performed for
1200 %      each right hand side.  This option determines the maximum number of
1201 %      iterative refinement steps.
1202 %
1203 % residual_ratio_max                     0 <  (      1e-10) <  +inf
1204 %    Iterative refinement tolerance
1205 %      Iterative refinement is performed until the residual test ratio is less
1206 %      than this tolerance (or until "max_refinement_steps" refinement steps are
1207 %      performed).
1208 %
1209 % residual_ratio_singular                0 <  (      1e-05) <  +inf
1210 %    Threshold for declaring linear system singular after failed iterative
1211 %    refinement.
1212 %      If the residual test ratio is larger than this value after failed
1213 %      iterative refinement, the algorithm pretends that the linear system is
1214 %      singular.
1215 %
1216 % residual_improvement_factor            0 <  (          1) <  +inf
1217 %    Minimal required reduction of residual test ratio in iterative refinement.
1218 %      If the improvement of the residual test ratio made by one iterative
1219 %      refinement step is not better than this factor, iterative refinement is
1220 %      aborted.
1221 %
1222 % neg_curv_test_tol                      0 <  (          0) <  +inf
1223 %    Tolerance for heuristic to ignore wrong inertia.
1224 %      If positive, incorrect inertia in the augmented system is ignored, and we
1225 %      test if the direction is a direction of positive curvature.  This
1226 %      tolerance determines when the direction is considered to be sufficiently
1227 %      positive.
1228 %
1229 % max_hessian_perturbation               0 <  (      1e+20) <  +inf
1230 %    Maximum value of regularization parameter for handling negative curvature.
1231 %      In order to guarantee that the search directions are indeed proper
1232 %      descent directions, Ipopt requires that the inertia of the (augmented)
1233 %      linear system for the step computation has the correct number of negative
1234 %      and positive eigenvalues. The idea is that this guides the algorithm away
1235 %      from maximizers and makes Ipopt more likely converge to first order
1236 %      optimal points that are minimizers. If the inertia is not correct, a
1237 %      multiple of the identity matrix is added to the Hessian of the Lagrangian
1238 %      in the augmented system. This parameter gives the maximum value of the
1239 %      regularization parameter. If a regularization of that size is not enough,
1240 %      the algorithm skips this iteration and goes to the restoration phase.
1241 %      (This is delta_w^max in the implementation paper.)
1242 %
1243 % min_hessian_perturbation               0 <= (      1e-20) <  +inf
1244 %    Smallest perturbation of the Hessian block.
1245 %      The size of the perturbation of the Hessian block is never selected
1246 %      smaller than this value, unless no perturbation is necessary. (This is
1247 %      delta_w^min in implementation paper.)
1248 %
1249 % perturb_inc_fact_first                 1 <  (        100) <  +inf
1250 %    Increase factor for x-s perturbation for very first perturbation.
1251 %      The factor by which the perturbation is increased when a trial value was
1252 %      not sufficient - this value is used for the computation of the very first
1253 %      perturbation and allows a different value for for the first perturbation
1254 %      than that used for the remaining perturbations. (This is bar_kappa_w^+ in
1255 %      the implementation paper.)
1256 %
1257 % perturb_inc_fact                       1 <  (          8) <  +inf
1258 %    Increase factor for x-s perturbation.
1259 %      The factor by which the perturbation is increased when a trial value was
1260 %      not sufficient - this value is used for the computation of all
1261 %      perturbations except for the first. (This is kappa_w^+ in the
1262 %      implementation paper.)
1263 %
1264 % perturb_dec_fact                       0 <  (   0.333333) <  1
1265 %    Decrease factor for x-s perturbation.
1266 %      The factor by which the perturbation is decreased when a trial value is
1267 %      deduced from the size of the most recent successful perturbation. (This
1268 %      is kappa_w^- in the implementation paper.)
1269 %
1270 % first_hessian_perturbation             0 <  (     0.0001) <  +inf
1271 %    Size of first x-s perturbation tried.
1272 %      The first value tried for the x-s perturbation in the inertia correction
1273 %      scheme.(This is delta_0 in the implementation paper.)
1274 %
1275 % jacobian_regularization_value          0 <= (      1e-08) <  +inf
1276 %    Size of the regularization for rank-deficient constraint Jacobians.
1277 %      (This is bar delta_c in the implementation paper.)
1278 %
1279 % jacobian_regularization_exponent         0 <= (       0.25) <  +inf
1280 %    Exponent for mu in the regularization for rank-deficient constraint
1281 %    Jacobians.
1282 %      (This is kappa_c in the implementation paper.)
1283 %
1284 % perturb_always_cd             ("no")
1285 %    Active permanent perturbation of constraint linearization.
1286 %      This options makes the delta_c and delta_d perturbation be used for the
1287 %      computation of every search direction.  Usually, it is only used when the
1288 %      iteration matrix is singular.
1289 %    Possible values:
1290 %     - no                      [perturbation only used when required]
1291 %     - yes                     [always use perturbation]
1292 %
1293 %
1294 %
1295 % ### Restoration Phase ###
1296 %
1297 % expect_infeasible_problem     ("no")
1298 %    Enable heuristics to quickly detect an infeasible problem.
1299 %      This options is meant to activate heuristics that may speed up the
1300 %      infeasibility determination if you expect that there is a good chance for
1301 %      the problem to be infeasible.  In the filter line search procedure, the
1302 %      restoration phase is called more quickly than usually, and more reduction
1303 %      in the constraint violation is enforced before the restoration phase is
1304 %      left. If the problem is square, this option is enabled automatically.
1305 %    Possible values:
1306 %     - no                      [the problem probably be feasible]
1307 %     - yes                     [the problem has a good chance to be infeasible]
1308 %
1309 % expect_infeasible_problem_ctol         0 <= (      0.001) <  +inf
1310 %    Threshold for disabling "expect_infeasible_problem" option.
1311 %      If the constraint violation becomes smaller than this threshold, the
1312 %      "expect_infeasible_problem" heuristics in the filter line search are
1313 %      disabled. If the problem is square, this options is set to 0.
1314 %
1315 % expect_infeasible_problem_ytol         0 <  (      1e+08) <  +inf
1316 %    Multiplier threshold for activating "expect_infeasible_problem" option.
1317 %      If the max norm of the constraint multipliers becomes larger than this
1318 %      value and "expect_infeasible_problem" is chosen, then the restoration
1319 %      phase is entered.
1320 %
1321 % start_with_resto              ("no")
1322 %    Tells algorithm to switch to restoration phase in first iteration.
1323 %      Setting this option to "yes" forces the algorithm to switch to the
1324 %      feasibility restoration phase in the first iteration. If the initial
1325 %      point is feasible, the algorithm will abort with a failure.
1326 %    Possible values:
1327 %     - no                      [don't force start in restoration phase]
1328 %     - yes                     [force start in restoration phase]
1329 %
1330 % soft_resto_pderror_reduction_factor         0 <= (     0.9999) <  +inf
1331 %    Required reduction in primal-dual error in the soft restoration phase.
1332 %      The soft restoration phase attempts to reduce the primal-dual error with
1333 %      regular steps. If the damped primal-dual step (damped only to satisfy the
1334 %      fraction-to-the-boundary rule) is not decreasing the primal-dual error by
1335 %      at least this factor, then the regular restoration phase is called.
1336 %      Choosing "0" here disables the soft restoration phase.
1337 %
1338 % max_soft_resto_iters                   0 <= (         10) <  +inf
1339 %    Maximum number of iterations performed successively in soft restoration
1340 %    phase.
1341 %      If the soft restoration phase is performed for more than so many
1342 %      iterations in a row, the regular restoration phase is called.
1343 %
1344 % required_infeasibility_reduction         0 <= (        0.9) <  1
1345 %    Required reduction of infeasibility before leaving restoration phase.
1346 %      The restoration phase algorithm is performed, until a point is found that
1347 %      is acceptable to the filter and the infeasibility has been reduced by at
1348 %      least the fraction given by this option.
1349 %
1350 % max_resto_iter                         0 <= (    3000000) <  +inf
1351 %    Maximum number of successive iterations in restoration phase.
1352 %      The algorithm terminates with an error message if the number of
1353 %      iterations successively taken in the restoration phase exceeds this
1354 %      number.
1355 %
1356 % evaluate_orig_obj_at_resto_trial("yes")
1357 %    Determines if the original objective function should be evaluated at
1358 %    restoration phase trial points.
1359 %      Setting this option to "yes" makes the restoration phase algorithm
1360 %      evaluate the objective function of the original problem at every trial
1361 %      point encountered during the restoration phase, even if this value is not
1362 %      required.  In this way, it is guaranteed that the original objective
1363 %      function can be evaluated without error at all accepted iterates;
1364 %      otherwise the algorithm might fail at a point where the restoration phase
1365 %      accepts an iterate that is good for the restoration phase problem, but
1366 %      not the original problem.  On the other hand, if the evaluation of the
1367 %      original objective is expensive, this might be costly.
1368 %    Possible values:
1369 %     - no                      [skip evaluation]
1370 %     - yes                     [evaluate at every trial point]
1371 %
1372 % resto_penalty_parameter                0 <  (       1000) <  +inf
1373 %    Penalty parameter in the restoration phase objective function.
1374 %      This is the parameter rho in equation (31a) in the Ipopt implementation
1375 %      paper.
1376 %
1377 % bound_mult_reset_threshold             0 <= (       1000) <  +inf
1378 %    Threshold for resetting bound multipliers after the restoration phase.
1379 %      After returning from the restoration phase, the bound multipliers are
1380 %      updated with a Newton step for complementarity.  Here, the change in the
1381 %      primal variables during the entire restoration phase is taken to be the
1382 %      corresponding primal Newton step. However, if after the update the
1383 %      largest bound multiplier exceeds the threshold specified by this option,
1384 %      the multipliers are all reset to 1.
1385 %
1386 % constr_mult_reset_threshold            0 <= (          0) <  +inf
1387 %    Threshold for resetting equality and inequality multipliers after
1388 %    restoration phase.
1389 %      After returning from the restoration phase, the constraint multipliers
1390 %      are recomputed by a least square estimate.  This option triggers when
1391 %      those least-square estimates should be ignored.
1392 %
1393 %
1394 %
1395 % ### Derivative Checker ###
1396 %
1397 % derivative_test               ("none")
1398 %    Enable derivative checker
1399 %      If this option is enabled, a (slow!) derivative test will be performed
1400 %      before the optimization.  The test is performed at the user provided
1401 %      starting point and marks derivative values that seem suspicious
1402 %    Possible values:
1403 %     - none                    [do not perform derivative test]
1404 %     - first-order             [perform test of first derivatives at starting
1405 %                                point]
1406 %     - second-order            [perform test of first and second derivatives at
1407 %                                starting point]
1408 %     - only-second-order       [perform test of second derivatives at starting
1409 %                                point]
1410 %
1411 % derivative_test_first_index           -2 <= (         -2) <  +inf
1412 %    Index of first quantity to be checked by derivative checker
1413 %      If this is set to -2, then all derivatives are checked.  Otherwise, for
1414 %      the first derivative test it specifies the first variable for which the
1415 %      test is done (counting starts at 0).  For second derivatives, it
1416 %      specifies the first constraint for which the test is done; counting of
1417 %      constraint indices starts at 0, and -1 refers to the objective function
1418 %      Hessian.
1419 %
1420 % derivative_test_perturbation           0 <  (      1e-08) <  +inf
1421 %    Size of the finite difference perturbation in derivative test.
1422 %      This determines the relative perturbation of the variable entries.
1423 %
1424 % derivative_test_tol                    0 <  (     0.0001) <  +inf
1425 %    Threshold for indicating wrong derivative.
1426 %      If the relative deviation of the estimated derivative from the given one
1427 %      is larger than this value, the corresponding derivative is marked as
1428 %      wrong.
1429 %
1430 % derivative_test_print_all     ("no")
1431 %    Indicates whether information for all estimated derivatives should be
1432 %    printed.
1433 %      Determines verbosity of derivative checker.
1434 %    Possible values:
1435 %     - no                      [Print only suspect derivatives]
1436 %     - yes                     [Print all derivatives]
1437 %
1438 % jacobian_approximation        ("exact")
1439 %    Specifies technique to compute constraint Jacobian
1440 %    Possible values:
1441 %     - exact                   [user-provided derivatives]
1442 %     - finite-difference-values [user-provided structure, values by finite
1443 %                                differences]
1444 %
1445 % findiff_perturbation                   0 <  (      1e-07) <  +inf
1446 %    Size of the finite difference perturbation for derivative approximation.
1447 %      This determines the relative perturbation of the variable entries.
1448 %
1449 % point_perturbation_radius              0 <= (         10) <  +inf
1450 %    Maximal perturbation of an evaluation point.
1451 %      If a random perturbation of a points is required, this number indicates
1452 %      the maximal perturbation.  This is for example used when determining the
1453 %      center point at which the finite difference derivative test is executed.
1454 %
1455 %
1456 %
1457 % ### Hessian Approximation ###
1458 %
1459 % limited_memory_max_history             0 <= (          6) <  +inf
1460 %    Maximum size of the history for the limited quasi-Newton Hessian
1461 %    approximation.
1462 %      This option determines the number of most recent iterations that are
1463 %      taken into account for the limited-memory quasi-Newton approximation.
1464 %
1465 % limited_memory_update_type    ("bfgs")
1466 %    Quasi-Newton update formula for the limited memory approximation.
1467 %      Determines which update formula is to be used for the limited-memory
1468 %      quasi-Newton approximation.
1469 %    Possible values:
1470 %     - bfgs                    [BFGS update (with skipping)]
1471 %     - sr1                     [SR1 (not working well)]
1472 %
1473 % limited_memory_initialization ("scalar1")
1474 %    Initialization strategy for the limited memory quasi-Newton approximation.
1475 %      Determines how the diagonal Matrix B_0 as the first term in the limited
1476 %      memory approximation should be computed.
1477 %    Possible values:
1478 %     - scalar1                 [sigma = s^Ty/s^Ts]
1479 %     - scalar2                 [sigma = y^Ty/s^Ty]
1480 %     - constant                [sigma = limited_memory_init_val]
1481 %
1482 % limited_memory_init_val                0 <  (          1) <  +inf
1483 %    Value for B0 in low-rank update.
1484 %      The starting matrix in the low rank update, B0, is chosen to be this
1485 %      multiple of the identity in the first iteration (when no updates have
1486 %      been performed yet), and is constantly chosen as this value, if
1487 %      "limited_memory_initialization" is "constant".
1488 %
1489 % limited_memory_init_val_max            0 <  (      1e+08) <  +inf
1490 %    Upper bound on value for B0 in low-rank update.
1491 %      The starting matrix in the low rank update, B0, is chosen to be this
1492 %      multiple of the identity in the first iteration (when no updates have
1493 %      been performed yet), and is constantly chosen as this value, if
1494 %      "limited_memory_initialization" is "constant".
1495 %
1496 % limited_memory_init_val_min            0 <  (      1e-08) <  +inf
1497 %    Lower bound on value for B0 in low-rank update.
1498 %      The starting matrix in the low rank update, B0, is chosen to be this
1499 %      multiple of the identity in the first iteration (when no updates have
1500 %      been performed yet), and is constantly chosen as this value, if
1501 %      "limited_memory_initialization" is "constant".
1502 %
1503 % limited_memory_max_skipping            1 <= (          2) <  +inf
1504 %    Threshold for successive iterations where update is skipped.
1505 %      If the update is skipped more than this number of successive iterations,
1506 %      we quasi-Newton approximation is reset.
1507 %
1508 % hessian_approximation         ("exact")
1509 %    Indicates what Hessian information is to be used.
1510 %      This determines which kind of information for the Hessian of the
1511 %      Lagrangian function is used by the algorithm.
1512 %    Possible values:
1513 %     - exact                   [Use second derivatives provided by the NLP.]
1514 %     - limited-memory          [Perform a limited-memory quasi-Newton
1515 %                                approximation]
1516 %
1517 % hessian_approximation_space   ("nonlinear-variables")
1518 %    Indicates in which subspace the Hessian information is to be approximated.
1519 %    Possible values:
1520 %     - nonlinear-variables     [only in space of nonlinear variables.]
1521 %     - all-variables           [in space of all variables (without slacks)]
1522 %
1523 %
1524 %
1525 % ### MA27 Linear Solver ###
1526 %
1527 % ma27_pivtol                            0 <  (      1e-08) <  1
1528 %    Pivot tolerance for the linear solver MA27.
1529 %      A smaller number pivots for sparsity, a larger number pivots for
1530 %      stability.  This option is only available if Ipopt has been compiled with
1531 %      MA27.
1532 %
1533 % ma27_pivtolmax                         0 <  (     0.0001) <  1
1534 %    Maximum pivot tolerance for the linear solver MA27.
1535 %      Ipopt may increase pivtol as high as pivtolmax to get a more accurate
1536 %      solution to the linear system.  This option is only available if Ipopt
1537 %      has been compiled with MA27.
1538 %
1539 % ma27_liw_init_factor                   1 <= (          5) <  +inf
1540 %    Integer workspace memory for MA27.
1541 %      The initial integer workspace memory = liw_init_factor * memory required
1542 %      by unfactored system. Ipopt will increase the workspace size by
1543 %      meminc_factor if required.  This option is only available if Ipopt has
1544 %      been compiled with MA27.
1545 %
1546 % ma27_la_init_factor                    1 <= (          5) <  +inf
1547 %    Real workspace memory for MA27.
1548 %      The initial real workspace memory = la_init_factor * memory required by
1549 %      unfactored system. Ipopt will increase the workspace size by
1550 %      meminc_factor if required.  This option is only available if  Ipopt has
1551 %      been compiled with MA27.
1552 %
1553 % ma27_meminc_factor                     1 <= (         10) <  +inf
1554 %    Increment factor for workspace size for MA27.
1555 %      If the integer or real workspace is not large enough, Ipopt will increase
1556 %      its size by this factor.  This option is only available if Ipopt has been
1557 %      compiled with MA27.
1558 %
1559 % ma27_skip_inertia_check       ("no")
1560 %    Always pretend inertia is correct.
1561 %      Setting this option to "yes" essentially disables inertia check. This
1562 %      option makes the algorithm non-robust and easily fail, but it might give
1563 %      some insight into the necessity of inertia control.
1564 %    Possible values:
1565 %     - no                      [check inertia]
1566 %     - yes                     [skip inertia check]
1567 %
1568 % ma27_ignore_singularity       ("no")
1569 %    Enables MA27's ability to solve a linear system even if the matrix is
1570 %    singular.
1571 %      Setting this option to "yes" means that Ipopt will call MA27 to compute
1572 %      solutions for right hand sides, even if MA27 has detected that the matrix
1573 %      is singular (but is still able to solve the linear system). In some cases
1574 %      this might be better than using Ipopt's heuristic of small perturbation
1575 %      of the lower diagonal of the KKT matrix.
1576 %    Possible values:
1577 %     - no                      [Don't have MA27 solve singular systems]
1578 %     - yes                     [Have MA27 solve singular systems]
1579 %
1580 %
1581 %
1582 % ### MA57 Linear Solver ###
1583 %
1584 % ma57_pivtol                            0 <  (      1e-08) <  1
1585 %    Pivot tolerance for the linear solver MA57.
1586 %      A smaller number pivots for sparsity, a larger number pivots for
1587 %      stability. This option is only available if Ipopt has been compiled with
1588 %      MA57.
1589 %
1590 % ma57_pivtolmax                         0 <  (     0.0001) <  1
1591 %    Maximum pivot tolerance for the linear solver MA57.
1592 %      Ipopt may increase pivtol as high as ma57_pivtolmax to get a more
1593 %      accurate solution to the linear system.  This option is only available if
1594 %      Ipopt has been compiled with MA57.
1595 %
1596 % ma57_pre_alloc                         1 <= (          3) <  +inf
1597 %    Safety factor for work space memory allocation for the linear solver MA57.
1598 %      If 1 is chosen, the suggested amount of work space is used.  However,
1599 %      choosing a larger number might avoid reallocation if the suggest values
1600 %      do not suffice.  This option is only available if Ipopt has been compiled
1601 %      with MA57.
1602 %
1603 % ma57_pivot_order                       0 <= (          5) <= 5
1604 %    Controls pivot order in MA57
1605 %      This is INCTL(6) in MA57.
1606 %
1607 %
1608 %
1609 % ### Pardiso Linear Solver ###
1610 %
1611 % pardiso_matching_strategy     ("complete+2x2")
1612 %    Matching strategy to be used by Pardiso
1613 %      This is IPAR(13) in Pardiso manual.  This option is only available if
1614 %      Ipopt has been compiled with Pardiso.
1615 %    Possible values:
1616 %     - complete                [Match complete (IPAR(13)=1)]
1617 %     - complete+2x2            [Match complete+2x2 (IPAR(13)=2)]
1618 %     - constraints             [Match constraints (IPAR(13)=3)]
1619 %
1620 % pardiso_redo_symbolic_fact_only_if_inertia_wrong("no")
1621 %    Toggle for handling case when elements were perturbed by Pardiso.
1622 %      This option is only available if Ipopt has been compiled with Pardiso.
1623 %    Possible values:
1624 %     - no                      [Always redo symbolic factorization when
1625 %                                elements were perturbed]
1626 %     - yes                     [Only redo symbolic factorization when elements
1627 %                                were perturbed if also the inertia was wrong]
1628 %
1629 % pardiso_repeated_perturbation_means_singular("no")
1630 %    Interpretation of perturbed elements.
1631 %      This option is only available if Ipopt has been compiled with Pardiso.
1632 %    Possible values:
1633 %     - no                      [Don't assume that matrix is singular if
1634 %                                elements were perturbed after recent symbolic
1635 %                                factorization]
1636 %     - yes                     [Assume that matrix is singular if elements were
1637 %                                perturbed after recent symbolic factorization]
1638 %
1639 % pardiso_out_of_core_power              0 <= (          0) <  +inf
1640 %    Enables out-of-core variant of Pardiso
1641 %      Setting this option to a positive integer k makes Pardiso work in the
1642 %      out-of-core variant where the factor is split in 2^k subdomains.  This is
1643 %      IPARM(50) in the Pardiso manual.  This option is only available if Ipopt
1644 %      has been compiled with Pardiso.
1645 %
1646 % pardiso_msglvl                         0 <= (          0) <  +inf
1647 %    Pardiso message level
1648 %      This determines the amount of analysis output from the Pardiso solver.
1649 %      This is MSGLVL in the Pardiso manual.
1650 %
1651 % pardiso_skip_inertia_check    ("no")
1652 %    Always pretent inertia is correct.
1653 %      Setting this option to "yes" essentially disables inertia check. This
1654 %      option makes the algorithm non-robust and easily fail, but it might give
1655 %      some insight into the necessity of inertia control.
1656 %    Possible values:
1657 %     - no                      [check inertia]
1658 %     - yes                     [skip inertia check]
1659 %
1660 % pardiso_max_iter                       1 <= (        500) <  +inf
1661 %    Maximum number of Krylov-Subspace Iteration
1662 %      DPARM(1)
1663 %
1664 % pardiso_iter_relative_tol              0 <  (      1e-06) <  1
1665 %    Relative Residual Convergence
1666 %      DPARM(2)
1667 %
1668 % pardiso_iter_coarse_size               1 <= (       5000) <  +inf
1669 %    Maximum Size of Coarse Grid Matrix
1670 %      DPARM(3)
1671 %
1672 % pardiso_iter_max_levels                1 <= (      10000) <  +inf
1673 %    Maximum Size of Grid Levels
1674 %      DPARM(4)
1675 %
1676 % pardiso_iter_dropping_factor           0 <  (        0.5) <  1
1677 %    dropping value for incomplete factor
1678 %      DPARM(5)
1679 %
1680 % pardiso_iter_dropping_schur            0 <  (        0.1) <  1
1681 %    dropping value for sparsify schur complement factor
1682 %      DPARM(6)
1683 %
1684 % pardiso_iter_max_row_fill              1 <= (   10000000) <  +inf
1685 %    max fill for each row
1686 %      DPARM(7)
1687 %
1688 % pardiso_iter_inverse_norm_factor         1 <  (      5e+06) <  +inf
1689 %
1690 %      DPARM(8)
1691 %
1692 % pardiso_iterative             ("no")
1693 %    Switch on iterative solver in Pardiso library
1694 %    Possible values:
1695 %     - no                      []
1696 %     - yes                     []
1697 %
1698 % pardiso_max_droptol_corrections         1 <= (          4) <  +inf
1699 %    Maximal number of decreases of drop tolerance during one solve.
1700 %      This is relevant only for iterative Pardiso options.
1701 %
1702 %
1703 %
1704 % ### Mumps Linear Solver ###
1705 %
1706 % mumps_pivtol                           0 <= (      1e-06) <= 1
1707 %    Pivot tolerance for the linear solver MUMPS.
1708 %      A smaller number pivots for sparsity, a larger number pivots for
1709 %      stability.  This option is only available if Ipopt has been compiled with
1710 %      MUMPS.
1711 %
1712 % mumps_pivtolmax                        0 <= (        0.1) <= 1
1713 %    Maximum pivot tolerance for the linear solver MUMPS.
1714 %      Ipopt may increase pivtol as high as pivtolmax to get a more accurate
1715 %      solution to the linear system.  This option is only available if Ipopt
1716 %      has been compiled with MUMPS.
1717 %
1718 % mumps_mem_percent                      0 <= (       1000) <  +inf
1719 %    Percentage increase in the estimated working space for MUMPS.
1720 %      In MUMPS when significant extra fill-in is caused by numerical pivoting,
1721 %      larger values of mumps_mem_percent may help use the workspace more
1722 %      efficiently.  On the other hand, if memory requirement are too large at
1723 %      the very beginning of the optimization, choosing a much smaller value for
1724 %      this option, such as 5, might reduce memory requirements.
1725 %
1726 % mumps_permuting_scaling                0 <= (          7) <= 7
1727 %    Controls permuting and scaling in MUMPS
1728 %      This is ICNTL(6) in MUMPS.
1729 %
1730 % mumps_pivot_order                      0 <= (          7) <= 7
1731 %    Controls pivot order in MUMPS
1732 %      This is ICNTL(7) in MUMPS.
1733 %
1734 % mumps_scaling                         -2 <= (         77) <= 77
1735 %    Controls scaling in MUMPS
1736 %      This is ICNTL(8) in MUMPS.
1737 %
1738 % mumps_dep_tol                       -inf <  (         -1) <  +inf
1739 %    Pivot threshold for detection of linearly dependent constraints in MUMPS.
1740 %      When MUMPS is used to determine linearly dependent constraints, this is
1741 %      determines the threshold for a pivot to be considered zero.  This is
1742 %      CNTL(3) in MUMPS.
1743 %
1744 %
1745 %
1746 % ### MA28 Linear Solver ###
1747 %
1748 % ma28_pivtol                            0 <  (       0.01) <= 1
1749 %    Pivot tolerance for linear solver MA28.
1750 %      This is used when MA28 tries to find the dependent constraints.
1751 %
1752 %
1753 %
1754 % ### Uncategorized ###
1755 %
1756 % warm_start_target_mu                -inf <  (          0) <  +inf
1757 %    Unsupported!

Generated on Fri 20-Mar-2015 18:23:34 by m2html © 2005