0

I'm using LPSolve IDE and have a problem with rounding in constraint matrix. For example,

min:; 1275699039.79*X1 + 1304079473.01209677*X2 <= 204401963493.5341; And LPSolve build matrix

      X1                        X2

R1: 1275699039.7903           1304079473.0121

Source multipler not equals matrix value(1275699039.79 != 1275699039.7903)

If change X1 multiplier to 111111275699039.79 then LPSolve matrix value = 111111275699053.

Is where any way to fix source multiplier?

A. Belkin
  • 78
  • 6
  • The magnitude of these coefficients is not appropriate for standard solvers. – Erwin Kalvelagen May 14 '18 at 08:13
  • @Erwin Kalvelagen, it works with small numbers too. Value 9039.79 has changed to 9039.789999999693 in matrix – A. Belkin May 14 '18 at 08:37
  • 1
    That is to be expected. This should not be a problem for normal models. Even if your input would be 100% correct, subsequent calculations would also deal with round off errors and tolerances. If your model assumes no rounding errors when dealing with floating point numbers you need a solver that does not use floating point numbers (there are a few rational solvers around). – Erwin Kalvelagen May 14 '18 at 08:44
  • @Erwin Kalvelagen, I would be appreciate for names of few of them. – A. Belkin May 14 '18 at 08:56
  • [QSopt_ex](https://www.math.uwaterloo.ca/~bico/qsopt/ex/index.html) and [SCIP/Soplex](http://scip.zib.de/) come to mind. – Erwin Kalvelagen May 14 '18 at 12:10

0 Answers0