use cplex for mixed integer problem but get non-integer output

Hello, i use cplex in my madlab code to solve a mixed integer problem. So I use the function cplexmilp(fu,Aineq,bineq,Aeq,beq,sostype,sosind,soswt,lb,ub,ctype), which should produce integer solutions because ctype only contains I and B. Most of the time it works fine but every now and then I get an "optimal solution" containing entries like -0.0000 or 1.0000 and the rest of my program breaks or gives the wrong output. Do you have any idea where my mistake might be? Thank you

Answer 1

[Alan Weiss]

Nearly any MILP solver uses floating point arithmetic internally, and uses tolerances to decide whether a number is "integer enough."

If you need really integer values for some variables, round them. If intcon is the vector of indices of the integer variables, execute

 x(intcon) = round(x(intcon));

or some such thing. See Some Integer Solutions Are Not Integers for a brief discussion of this issue related to the MATLAB intlinprog solver.

Alan Weiss

MATLAB mathematical toolbox documentation