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

151 lines
4.2 KiB
Matlab
Executable File

%%******************************************************************
%% symqmr: symmetric QMR with left (symmetric) preconditioner.
%% The preconditioner used is based on the analytical
%% expression of inv(A).
%%
%% [x,resnrm,solve_ok] = symqmr(A,b,L,tol,maxit)
%%
%% child function: linsysolvefun.m
%%
%% A = [mat11 mat12; mat12' mat22].
%% b = rhs vector.
%% if matfct_options = 'chol' or 'spchol'
%% L = Cholesky factorization of (1,1) block.
%% M = Cholesky factorization of
%% Schur complement of A ( = mat12'*inv(mat11)*mat12-mat22).
%% else
%% L = triangular factors of A.
%% M = not relevant.
%% end
%% resnrm = norm of qmr-generated residual vector b-Ax.
%%*****************************************************************
%% SDPT3: version 4.0
%% Copyright (c) 1997 by
%% Kim-Chuan Toh, Michael J. Todd, Reha H. Tutuncu
%% Last Modified: 16 Sep 2004
%%*****************************************************************
function [xx,resnrm,solve_ok] = symqmr(A,b,L,tol,maxit,printlevel)
N = length(b);
if (nargin < 6); printlevel = 1; end
if (nargin < 5) | isempty(maxit); maxit = max(30,length(A.mat22)); end;
if (nargin < 4) | isempty(tol); tol = 1e-10; end;
tolb = min(1e-4,tol*norm(b));
solve_ok = 1;
x = zeros(N,1);
if (norm(x))
if isstruct(A); Aq = matvec(A,x); else; Aq=A*x; end;
r = b-Aq;
else
r = b;
end
err = norm(r); resnrm(1) = err; minres = err; xx = x;
if (err < 1e-3*tolb); return; end
q = precond(A,L,r);
tau_old = norm(q);
rho_old = r'*q;
theta_old = 0;
d = zeros(N,1);
res = r; Ad = zeros(N,1);
%%
%% main loop
%%
tiny = 1e-30;
for iter = 1:maxit
if isstruct(A); Aq = matvec(A,q); else; Aq=A*q; end;
sigma = q'*Aq;
if (abs(sigma) < tiny)
solve_ok = 2;
if (printlevel); fprintf('*'); end;
break;
else
alpha = rho_old/sigma;
r = r - alpha*Aq;
end
u = precond(A,L,r);
%%
theta = norm(u)/tau_old; c = 1/sqrt(1+theta^2);
tau = tau_old*theta*c;
gam = (c^2*theta_old^2); eta = (c^2*alpha);
d = gam*d + eta*q;
x = x + d;
%%
Ad = gam*Ad + eta*Aq;
res = res - Ad;
err = norm(res); resnrm(iter+1) = err;
if (err < minres); xx = x; minres = err; end
if (err < tolb); break; end
if (iter > 10)
if (err > 0.98*mean(resnrm(iter-10:iter)))
solve_ok = 0.5; break;
end
end
%%
if (abs(rho_old) < tiny)
solve_ok = 2;
if (printlevel); fprintf('*'); end;
break;
else
rho = r'*u;
beta = rho/rho_old;
q = u + beta*q;
end
rho_old = rho;
tau_old = tau;
theta_old = theta;
end
if (iter == maxit); solve_ok = 0.3; end;
%%
%%*************************************************************************
%% precond:
%%*************************************************************************
function Mx = precond(A,L,x)
m = L.matdim; m2 = length(x)-m;
Mx = zeros(length(x),1);
for iter = 1:1
if norm(Mx); r = full(x - matvec(A,Mx)); else; r = full(x); end
r1 = r(1:m);
if (m2 > 0)
r2 = r(m+[1:m2]);
w = linsysolvefun(L,r1);
z = mexMatvec(A.mat12,w,1) - r2;
z = L.Mu \ (L.Ml \ (L.Mp*z));
r1 = r1 - mexMatvec(A.mat12,z);
end
d = linsysolvefun(L,r1);
if (m2 > 0)
d = [d; z];
end
Mx = Mx + d;
end
%%*************************************************************************
%% matvec: matrix-vector multiply.
%% matrix = [A.mat11, A.mat12; A.mat12', A.mat22]
%%*************************************************************************
function Ax = matvec(A,x);
m = length(A.mat11); m2 = length(x)-m;
if issparse(x); x = full(x); end
if (m2 > 0)
x1 = x(1:m);
else
x1 = x;
end
Ax = mexMatvec(A.mat11,x1);
if (m2 > 0)
x2 = x(m+[1:m2]);
Ax = Ax + mexMatvec(A.mat12,x2);
Ax2 = mexMatvec(A.mat12,x1,1) + mexMatvec(A.mat22,x2);
Ax = [full(Ax); full(Ax2)];
end
%%*************************************************************************