117 lines
3.4 KiB
Mathematica
117 lines
3.4 KiB
Mathematica
|
function pen = sedumi2pen(F_struc,K,c,x0)
|
||
|
|
%SEDUMI2MPEN Internal function to convert SeDuMi structure to format needed in PENNON
|
||
|
|
|
||
|
|
% General data
|
||
|
|
pen.vars = length(c);
|
||
|
|
pen.constr = K.l;
|
||
|
|
if length(K.s)>1
|
||
|
|
pen.mconstr = length(K.s);
|
||
|
|
else
|
||
|
|
if K.s==0
|
||
|
|
pen.mconstr=0;
|
||
|
|
else
|
||
|
|
pen.mconstr=1;
|
||
|
|
end
|
||
|
|
end
|
||
|
|
pen.msizes = K.s;
|
||
|
|
pen.fobj = c(:)';
|
||
|
|
pen.x0 = zeros(1,length(c));
|
||
|
|
|
||
|
|
% Linear constraints
|
||
|
|
if K.l>0
|
||
|
|
pen.ci = full(F_struc(K.f+1:K.f+K.l,1))';
|
||
|
|
pen.bi_dim = zeros(1,K.f+K.l);
|
||
|
|
[variables,constr,val] = find(-F_struc(1:K.l,2:end)');
|
||
|
|
pen.bi_idx = variables(:)'-1;
|
||
|
|
pen.bi_val = val(:)';
|
||
|
|
j = 1;
|
||
|
|
|
||
|
|
if ~isempty(constr)
|
||
|
|
rr = histc(constr,[1:max(constr)]);
|
||
|
|
uu = find(rr);
|
||
|
|
pen.bi_dim(uu) = rr(uu);
|
||
|
|
%uns = uniquestripped(constr);
|
||
|
|
%for j = 1:length(uns)
|
||
|
|
% pen.bi_dim(uns(j)) = sum(constr==uns(j));
|
||
|
|
%end
|
||
|
|
end
|
||
|
|
|
||
|
|
else
|
||
|
|
pen.ci = 0;
|
||
|
|
pen.bi_dim = 0;
|
||
|
|
pen.bi_idx = 0;
|
||
|
|
pen.bi_val = 0;
|
||
|
|
end
|
||
|
|
|
||
|
|
% Semidefnite constraints
|
||
|
|
if K.s>0
|
||
|
|
top = K.l+K.f+1;
|
||
|
|
constraints = 1;
|
||
|
|
pen.ai_dim = zeros(1,length(K.s));
|
||
|
|
|
||
|
|
% First, optimized code for scalar SDPs (happens for nonlinear scalar constraints)
|
||
|
|
if all(K.s==1)
|
||
|
|
|
||
|
|
Avec = -F_struc(top:top+length(K.s)-1,:);
|
||
|
|
[variables,constr,val] = find(Avec');
|
||
|
|
j = 1;
|
||
|
|
n_nz = length(val);
|
||
|
|
pen.ai_nzs = repmat(1,1,n_nz);
|
||
|
|
pen.ai_row = repmat(0,1,n_nz);
|
||
|
|
pen.ai_col = pen.ai_row;%repmat(0,1,n_nz);
|
||
|
|
pen.ai_val = val(:)';
|
||
|
|
pen.ai_idx = variables(:)'-1;
|
||
|
|
while j<=length(val)
|
||
|
|
first_constraint = constr(j);
|
||
|
|
n_nz = 1;
|
||
|
|
while j+n_nz<=length(val) & constr(j+n_nz)==first_constraint
|
||
|
|
n_nz = n_nz + 1;
|
||
|
|
end
|
||
|
|
pen.ai_dim(first_constraint) = n_nz;
|
||
|
|
j = j + n_nz;
|
||
|
|
end
|
||
|
|
else
|
||
|
|
pen.ai_idx = [];
|
||
|
|
pen.ai_val = [];
|
||
|
|
pen.ai_row = [];
|
||
|
|
pen.ai_col = [];
|
||
|
|
pen.ai_nzs = [];
|
||
|
|
while (constraints<=length(K.s))
|
||
|
|
n = K.s(constraints);
|
||
|
|
Avec = -F_struc(top:top+n^2-1,:);
|
||
|
|
picker = reshape(1:n^2,n,n);picker = tril(picker-diag(diag(picker)));picker = find(picker(:));
|
||
|
|
[rowcols,varindicies,vals]=find(Avec);
|
||
|
|
%the_col = 1+floor((rowcols-1)/n);
|
||
|
|
%the_row = rowcols-(the_col-1)*n;
|
||
|
|
%removethese = the_row > the_col;
|
||
|
|
removethese = find(ismembc(rowcols,picker));
|
||
|
|
rowcols(removethese) =[];
|
||
|
|
varindicies(removethese)=[];
|
||
|
|
vals(removethese)=[];
|
||
|
|
if ~isempty(rowcols)
|
||
|
|
cols = ceil(rowcols/n);
|
||
|
|
rows = rowcols - n*(cols-1);
|
||
|
|
uns = uniquestripped(varindicies);
|
||
|
|
|
||
|
|
difference = diff([varindicies(:) ; max(varindicies)+1]);
|
||
|
|
count = diff(find([1 ; difference]))';
|
||
|
|
pen.ai_nzs = [pen.ai_nzs count];
|
||
|
|
|
||
|
|
pen.ai_dim(constraints) = length(uniquestripped(varindicies));
|
||
|
|
pen.ai_idx = [pen.ai_idx uns(:)'-1];
|
||
|
|
pen.ai_row = [pen.ai_row rows(:)'-1];
|
||
|
|
pen.ai_col = [pen.ai_col cols(:)'-1];
|
||
|
|
pen.ai_val = [pen.ai_val vals(:)'];
|
||
|
|
end
|
||
|
|
constraints = constraints+1;
|
||
|
|
top = top+n*n;
|
||
|
|
end
|
||
|
|
end
|
||
|
|
else
|
||
|
|
pen.ai_dim = 0;
|
||
|
|
pen.ai_idx = 0;
|
||
|
|
pen.ai_val = 0;
|
||
|
|
pen.ai_row = 0;
|
||
|
|
pen.ai_col = 0;
|
||
|
|
pen.ai_nzs = 0;
|
||
|
|
end
|