How to set good bounds?
Posted: Wed Jan 30, 2019 10:35 am
Dear GAMS friends,
I use CONOPT to solve nonlinear economic models. I found bounds are very important for the model to be solved. If a variable can get a good bound, the model can be solved, and if cant, the model becomes infeasible. The CONOPT Solver manual also suggests to set bounds, but also points out the risk of redundent bounds. I am not clear when the bound is redundent, or which variable should not have bounds. Does a variable need to have a lower bound and an upper bound at the same time? And is there a general rule for a good range of bounds, say 1.E-6~1.E+10 for positive variables?
Many thanks,
Qiaomin
I use CONOPT to solve nonlinear economic models. I found bounds are very important for the model to be solved. If a variable can get a good bound, the model can be solved, and if cant, the model becomes infeasible. The CONOPT Solver manual also suggests to set bounds, but also points out the risk of redundent bounds. I am not clear when the bound is redundent, or which variable should not have bounds. Does a variable need to have a lower bound and an upper bound at the same time? And is there a general rule for a good range of bounds, say 1.E-6~1.E+10 for positive variables?
Many thanks,
Qiaomin