[Symphony] solving a sequence of MILPs using warm start feature

Cyrus Angelo Selga cselga2 at gmail.com
Thu May 14 04:32:39 EDT 2009


Hi, I'm currently using Symphony v5.1.8 to experiment with solving a 
sequence of MILPs using the warm start feature. The MILPs are specified 
using the MPS format and differ only on the RHS of the constraints. 
According  to the user manual, 
http://www.coin-or.org/SYMPHONY/man-5.1/node79.html, Symphony can handle 
changes to rim vectors.

However, after solving the first MILP and attempting to use the warm 
start info for the 2nd MILP, Symphony immediately terminates and 
declares the optimal solution of the first MILP. Here's the code which 
I'm trying to work on:

int main(int argc, char **argv)
{
   sym_environment *env = sym_open_environment();
   warm_start_desc * ws;
   sym_set_int_param(env, "verbosity", -1);

   sym_read_mps(env, "iter3.mps");
   sym_set_int_param(env, "keep_warm_start", 1);
   sym_solve(env);

   ws = sym_get_warm_start(env, true);
   sym_read_mps(env, "iter4.mps");
   sym_set_warm_start(env, ws);
   sym_warm_solve(env);
   
   sym_close_environment(env);
}

My problem seems to be similar to a previous post, 
http://list.coin-or.org/pipermail/symphony/2009-May/000448.html, though 
I'm not sure. Can the problem modification be handled by loading 
different MPS files or should they be explicit set by one of the data 
modification functions, e.g. sym_set_obj_coeff?

By the way, I'd like to thank the Symphony developers for making this 
software available. I'm currently a M.S. OR candidate and the COIN-OR 
software prove to be very valuable in my research.

Cyrus




More information about the Symphony mailing list