You are on page 1of 1

# Corollary 2 (Farkass Theorem as a Consequence of Theorem 2.7.

3)

Let A be an m n matrix and let c be an n-vector. Then exactly one of the following two

## System 2: At y = c and y 0 for some y R m .

Proof

{ }
Consider the linear program P to minimize 0t y : At y = c, y 0 . The dual D to this problem is

{
given by maximize ct x : Ax 0 . }
If System 1 has a solution, there exists x R n such that Ax 0 and ct x > 0 , where {x : Ax 0}

is a cone and {x : c t x > 0} . A solution x R n to System 1 lies in the cone {x : Ax 0} and the

cone {x : c t x > 0} as well. Therefore, the set of feasible solutions is the intersection of two cones,

{x : Ax 0} and {x : c x > 0} .
t
We also know the intersection of two cones, {x : Ax 0} and

{x : c x > 0} , is unbounded.
t

D is unbounded.
P is infeasible (because of 2.7.3 (c))
System 2 has no solution
If System 2 has a solution,
P has a solution y with the optimal objective value 0.
D has a solution with the optimal objective value 0 (because of Strong duality)
ct x 0 for all x with Ax 0 (the optimal objective value is zero so the objective values of all
feasible solutions 0 .)
System 1 has no solution.

27