CONOPT solver_result_num
2 views
Skip to first unread message
gabe.ha...@gmail.com
Jun 24, 2016, 12:15:55 PM6/24/16
to AMPL Modeling Language
When CONOPT is used to solve an NLP, the expected normal execution is that it returns a locally optimal solution. However, in this case, the solver_results_num (as documented here) is set to 100 (documented here).
Should this be a lower number, say in the range 0-99? According to the description of the solver_results_num ranges, 100-199 is reserved for cases where "optimal solution indicated, but error likely". Since CONOPT is not a global optimizer, returning a locally optimal solution in the case where convexity can not be proven seems like normal solver execution.
Gabe
Should this be a lower number, say in the range 0-99? According to the description of the solver_results_num ranges, 100-199 is reserved for cases where "optimal solution indicated, but error likely". Since CONOPT is not a global optimizer, returning a locally optimal solution in the case where convexity can not be proven seems like normal solver execution.
Gabe
Reply all
Reply to author
Forward
0 new messages