Why intlinprog cant solve 6561 variable optimization?

Hi all, I have a digraph of 81 nodes and I need to find the shortest path from a starting node to all the rest using linear programming. I adopted the idea from this post. But since I wanted from one node to all I set flow of all other nodes to be -1 and the starting node's flow as | V|-1. Adjacency matrix (named B) has size of 81x81 and values >0 if an edge exists between those nodes or Inf if not.

So i proceeded as follows:

V = size(B,1); % V = 81
flow = ones(1,V);
flow = -flow;
flow(startpoint) = V-1;

coeff = zeros(V,V,V);
for i=(1:V)
    for j=(1:V)
        coeff(i,i,j) = -1;
        coeff(i,j,i) = 1;
    end
    coeff(i,i,i) = 0;
end

intlinprog(B,81*81,[],[],coeff,flow,zeros(1,V^2),ones(1,V^2));

I think this is correct from the documentation although I'm not that sure. But the problem is that when I run it I get this error:

Error using intlinprog (line 142)
INTLINPROG stopped because some objective or constraint matrix coefficients are too large in magnitude.

[Matt J]

Hmmm. Well, I suggest you attach adjacencymatrix in a .mat file so that we can try it ourselves.

[asdasd asdasdas]

Done!

Answer 1

[Alan Weiss]

intlinprog restricts coefficients to no more than 1e25. So you cannot formulate things exactly the way you want. Perhaps try 1e20 instead of Inf.

Alan Weiss

MATLAB mathematical toolbox documentation

[asdasd asdasdas]

Thank you this will send me towards the right solution to my problem. Apparently since it's my first time dealing with so many dimensions I thought that was the problem so I ignored the part before 'or' in the error.