Originally posted by: BerkUstun

I remember learning about a tweaked version of the Simplex algorithm that accounted for bounded variables in my mathematical programming classes - so that users would get a slight computational benefit in solving an LP because constraints of the type x > lb or x < ub did not have to be tacked onto the Ax < b part of the problem.

Since the CPLEX API allows us to specify lb and ub parameters in Model.lb and Model.ub, does it use a similar algorithm when a user specifies lb/ub in the Model? Or will these constraints be tacked onto Model.A? Simply put, is there a computational benefit in specifying ub and lb in CPLEX?
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: John Cui

Which API are you using?

John Cui
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: BerkUstun

The MATLAB API
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: John Cui

First, we should split lb and ub for decision variables from A matrix, A matrix only handle constraints, like x1 + x2 - x3 <= 10.
Then it is more clear to know what constraints and bounds are.

But you can use A matrix to represent lb and ub as well. In theory, 1 <= x2 <= 10 is a constraint.

And I think it is good for performance if you specify the bounds of decision variables.

John Cui
#CPLEXOptimizers
#DecisionOptimization