testlp

testlp - inds=find(ds<0);

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
profile on clear all %%% Generate random data for an LP in SEF %%% with strictly feasible starting points. m=10;n=19; % dimensions of problem toler=1e-8; disp(['dimension m,n of problem is: ',num2str([m n])]) disp(['tolerance for duality gap is: ',num2str(toler)]) A=[rand(m,n-m) eye(m)];b=10*(rand(m,1)+1);c=rand(n,1)+1; x=ones(n,1);s=x;y=ones(m,1); b=A*x;c=A'*y-s; % ensure feasible start iters=0; while sum(x.*s) > toler, iters=iters+1; %% Find affine scaling search direction [dx,dy,ds]=intpteff(A,b,c,x,y,s); [dx2,dy2,ds2]=intpt(A,b,c,x,y,s); dir2=[dx2;dy2;ds2]; dir=[dx;dy;ds]; inds=find(dx<0); alphamin=min(-x(inds)./dx(inds));
Background image of page 1
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: inds=find(ds&lt;0); alphamin=min(alphamin,min(-s(inds)./ds(inds))); %%% Backtrack if necessary alpha=1; if alphamin &lt; 1, while min( min(x+alpha*dx),min(s+alpha*ds) ) &lt;= 0, alpha = .64*alphamin; end end normerror=norm(dir-dir2); disp(['iteration; gap, step , direction error : ',. .. num2str([iters sum(x.*s) alpha normerror])]); %% update x=x+alpha*dx;s=s+alpha*ds; y=y+alpha*dy; end disp(' ') disp(['number of iterations is: ',num2str(iters)]); disp(' ') disp(['optimal x s to duality gap: ',num2str(sum(x.*s))]) %[x s] profile report...
View Full Document

Ask a homework question - tutors are online