Dynamic-Calibration/utils/SDPT3-4.0/Solver/sqlparameters.m

92 lines
5.2 KiB
Matlab
Executable File

%%*************************************************************************
%% sqlparameters.m: set OPTIONS structure to specify default
%% parameters for sqlp.m
%%
%% OPTIONS.vers : version of direction to use.
%% 1 for HKM direction
%% 2 for NT direction
%% 0 for the default (which uses the HKM direction if
%% semidefinite blocks exist; and NT direction of SOCP problms)
%% OPTIONS.gam : step-length parameter,
%% OPTIONS.predcorr : whether to use Mehrotra predictor-corrector.
%% OPTIONS.expon : exponent in decrease of centering parameter sigma.
%% OPTIONS.gaptol : tolerance for duality gap as a fraction of the
%% value of the objective functions.
%% OPTIONS.inftol : tolerance for stopping due to suspicion of
%% infeasibility.
%% OPTIONS.steptol : toloerance for stopping due to small steps.
%% OPTIONS.maxit : maximum number of iteration allowed
%% OPTIONS.printlevel : 3, if want to display result in each iteration,
%% 2, if want to display only summary,
%% 1, if want to display warning message,
%% 0, no display at all.
%% OPTIONS.stoplevel : 2, if want to automatically detect termination;
%% 1, if want to automatically detect termination, but
%% restart automatically with a new iterate
%% when the algorithm stagnants because of tiny step-lengths.
%% 0, if want the algorithm to continue forever except for
%% successful completion, maximum number of iterations, or
%% numerical failures. Note, do not change this field unless
%% you very sure.
%% OPTIONS.scale_data : 1, if want to scale the data before solving the problem,
%% else = 0
%% OPTIONS.rmdepconstr : 1, if want to remove nearly dependent constraints,
%% else = 0.
%% OPTIONS.smallblkdim : block-size threshold determining what method to compute the
%% schur complement matrix corresponding to semidefintie block.
%% NOTE: this number should be small, say less than 20.
%% OPTIONS.parbarrier : parameter values of the log-barrier terms in the SQLP problem.
%% Default = [], meaning that the parameter values are all 0.
%% OPTIONS.schurfun : [], if no user supplied routine to compute the Schur matrix,
%% else, it is a cell array where each cell is either [],
%% or contains a string that is the file name where the Schur matrix
%% of the associated block data is computed.
%% For example, if the SQLP data has the block structure
%% blk{1,1} = '1'; blk{1,2} = 10;
%% blk{2,1} = 's'; blk{2,2} = 50;
%% and
%% OPTIONS.schurfun{1} = [];
%% OPTIONS.schurfun{2} = 'myownschur', where
%% 'myownschur' is a function with the calling sequence:
%% function schur = myownschur(X2,Z2inv,schurfun_par(2,:));
%% This means that for the first block, the Schur
%% matrix is computed by the default method in SDPT3,
%% and for the second block, the user supplies the
%% routine to compute the associated Schur matrix.
%% OPTIONS.schurfun_par: [], if no user supplied routine to compute the Schur matrix,
%% else, it is a cell array where the p-th row is either [],
%% or is a cell array containing the parameters needed in
%% the user supplied Schur routine OPTIONS.schurfun{p}.
%% For example, for the block structure described
%% above, we may have:
%% OPTIONS.schurfun_par{1} = [];
%% OPTIONS.schurfun_par{2,1} = par1;
%% OPTIONS.schurfun_par{2,2} = par2;
%%*****************************************************************
%% SDPT3: version 4.0
%% Copyright (c) 1997 by
%% Kim-Chuan Toh, Michael J. Todd, Reha H. Tutuncu
%% Last Modified: 16 Sep 2004
%%*****************************************************************
function OPTIONS = sqlparameters
OPTIONS.vers = 0;
OPTIONS.gam = 0;
OPTIONS.predcorr = 1;
OPTIONS.expon = 1;
OPTIONS.gaptol = 1e-8;
OPTIONS.inftol = 1e-8;
OPTIONS.steptol = 1e-6;
OPTIONS.maxit = 100;
OPTIONS.printlevel = 3;
OPTIONS.stoplevel = 1; %% do not change this field unless you very sure.
OPTIONS.scale_data = 0;
OPTIONS.spdensity = 0.4;
OPTIONS.rmdepconstr = 0;
OPTIONS.smallblkdim = 50;
OPTIONS.parbarrier = [];
OPTIONS.schurfun = [];
OPTIONS.schurfun_par = [];
%%*************************************************************************