# of Constraints in (D) = # of Variables in (P) Coefficients of Obj. Fn. In (D) comes from r.h.s. of constraints in (P) If (P) is a max model, then (D) is a min model. If (P) is a min model, then (D) is a max model Coefficients in the ith constraint function in (D) are the coefficients of the ith variable of constraints in (P) The r.h.ss of constraints in (D) come from the objective function coefficients in (P) The sense of the ith constraint in (D) is = iff the ith variable in (P) is unrestricted If (P) is a MAX problem, then the constraints in (D) have the same sense as the corresponding variables in (P). If (P) is a MIN problem, then the dual constraints have a sense opposite to the corresponding variable in the original problem The ith variable in (D) is unconstrained iff the ith constraint in (P) is an equality If (P) is a MAX problem, then the variables in (D) have sense opposite to the sign of the corresponding constraints in (P). If (P) is a MIN problem, then the dual variables have the same sense as the corresponding constraints in (P)
Weak Duality Theorem
Let X denote a feasible solution to (P), and Y a feasible solution to (D) If (P) is a MAX problem and (D) a MIN problem, then Obj.Val.(X) <= Obj.Val.(Y) If (P) is a MIN problem and (D) a MAX problem, then Obj.Val.(X) >= Obj.Val.(Y) Strong Duality Theorem: If (P) has a finite optimal solution then (D) also has a finite optimal solution, and vice versa. Further, in such a case the optimal objective function values for both problems are equal. That is, O.V.(X*) = O.V.(Y*) Principle of Complementary Slackness: At optimality, if ith constraint in (P) is loose, then the corresponding optimal dual variable Y*i = 0 At optimality, if Y*i > 0, then the corresponding primal constraint is tight.