105 lines
2.8 KiB
Mathematica
105 lines
2.8 KiB
Mathematica
|
function pen = sedumi2penbmi(F_struc,c,Q,K,monomtable,options,x0)
|
||
|
|
|
||
|
|
% Linear before bilinearize
|
||
|
|
temp = sum(monomtable,2)>1;
|
||
|
|
tempnonlinearindicies = find(temp);
|
||
|
|
templinearindicies = find(~temp);
|
||
|
|
|
||
|
|
if isempty(monomtable)
|
||
|
|
monomtable = eye(length(c));
|
||
|
|
end
|
||
|
|
|
||
|
|
temp = sum(monomtable,2)>1;
|
||
|
|
nonlinearindicies = find(temp);
|
||
|
|
linearindicies = find(~temp);
|
||
|
|
c0 = c;
|
||
|
|
c = c(linearindicies);
|
||
|
|
Q = Q(linearindicies,linearindicies);
|
||
|
|
nonlinear_scalars = [];
|
||
|
|
|
||
|
|
% Any non-linear scalar inequalities?
|
||
|
|
% Move these to the BMI part
|
||
|
|
if K.l>0
|
||
|
|
nonlinear_scalars = find(any(full(F_struc(1:K.l,[nonlinearindicies(:)'+1])),2));
|
||
|
|
if ~isempty(nonlinear_scalars)
|
||
|
|
Kold = K;
|
||
|
|
linear_scalars = setdiff(1:K.l,nonlinear_scalars);
|
||
|
|
F_struc = [F_struc(linear_scalars,:);F_struc(nonlinear_scalars,:);F_struc(K.l+1:end,:)];
|
||
|
|
K.l = K.l-length(nonlinear_scalars);
|
||
|
|
if (length(K.s)==1) & (K.s==0)
|
||
|
|
K.s = [repmat(1,1,length(nonlinear_scalars))];
|
||
|
|
K.rank = repmat(1,1,length(nonlinear_scalars));
|
||
|
|
else
|
||
|
|
K.s = [repmat(1,1,length(nonlinear_scalars)) K.s];
|
||
|
|
K.rank = [repmat(1,1,length(nonlinear_scalars)) K.rank];
|
||
|
|
end
|
||
|
|
end
|
||
|
|
end
|
||
|
|
|
||
|
|
if ~isempty(F_struc)
|
||
|
|
pen = sedumi2pen(F_struc(:,[1 linearindicies(:)'+1]),K,c,x0);
|
||
|
|
else
|
||
|
|
pen = sedumi2pen([],K,c,x0);
|
||
|
|
end
|
||
|
|
|
||
|
|
if ~isempty(nonlinearindicies)
|
||
|
|
bmi = sedumi2pen(F_struc(:,[nonlinearindicies(:)'+1]),K,[],[]);
|
||
|
|
pen.ki_dim = bmi.ai_dim;
|
||
|
|
% Nonlinear index
|
||
|
|
pen.ki_dim = bmi.ai_dim;
|
||
|
|
pen.ki_row = bmi.ai_row;
|
||
|
|
pen.ki_col = bmi.ai_col;
|
||
|
|
pen.ki_nzs = bmi.ai_nzs;
|
||
|
|
pen.ki_val = bmi.ai_val;
|
||
|
|
for i = 1:length(bmi.ai_idx)
|
||
|
|
nl = nonlinearindicies(1+bmi.ai_idx(i));
|
||
|
|
v = find(monomtable(nl,:));
|
||
|
|
if length(v)==1
|
||
|
|
v(2)=v(1);
|
||
|
|
end
|
||
|
|
pen.ki_idx(i)=find(linearindicies == v(1));
|
||
|
|
pen.kj_idx(i)=find(linearindicies == v(2));
|
||
|
|
end
|
||
|
|
else
|
||
|
|
pen.ki_dim = 0*pen.ai_dim;
|
||
|
|
pen.ki_row = 0;
|
||
|
|
pen.ki_col = 0;
|
||
|
|
pen.ki_nzs = 0;
|
||
|
|
pen.ki_idx = 0;
|
||
|
|
pen.kj_idx = 0;
|
||
|
|
pen.kj_val = 0;
|
||
|
|
end
|
||
|
|
|
||
|
|
if nnz(Q)>0
|
||
|
|
[row,col,vals] = find(triu(Q));
|
||
|
|
pen.q_nzs = length(row);
|
||
|
|
pen.Q_nzs = length(row);%Bug in PENLABs interface?
|
||
|
|
pen.q_val = vals';
|
||
|
|
pen.q_col = col'-1;
|
||
|
|
pen.q_row = row'-1;
|
||
|
|
else
|
||
|
|
pen.q_nzs = 0;
|
||
|
|
pen.q_val = 0;
|
||
|
|
pen.q_col = 0;
|
||
|
|
pen.q_row = 0;
|
||
|
|
end
|
||
|
|
|
||
|
|
ops = struct2cell(options.penbmi);ops = [ops{1:end}];
|
||
|
|
pen.ioptions = ops(1:12);
|
||
|
|
pen.foptions = ops(13:end);
|
||
|
|
pen.ioptions(4) = max(0,min(3,options.verbose+1));
|
||
|
|
if pen.ioptions(4)==1
|
||
|
|
pen.ioptions(4)=0;
|
||
|
|
end
|
||
|
|
|
||
|
|
% ****************************************
|
||
|
|
% UNCOMMENT THIS FOR PENBMI version 1
|
||
|
|
% ****************************************
|
||
|
|
%pen.ioptions = pen.ioptions(1:8);
|
||
|
|
%pen.foptions = pen.foptions(1:8);
|
||
|
|
|
||
|
|
if ~isempty(x0)
|
||
|
|
pen.x0(isnan(pen.x0)) = 0;
|
||
|
|
pen.x0 = x0(linearindicies);
|
||
|
|
pen.x0 = pen.x0(:)';
|
||
|
|
end
|