Linear Programming Lecture Notes - Penn State Personal Web Server
Linear Programming Lecture Notes - Penn State Personal Web Server
Linear Programming Lecture Notes - Penn State Personal Web Server
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
If there is some i such that c T di > 0, then we can simply choose µi as large as we like,<br />
making the objective as large as we like, the problem will have no finite solution.<br />
Therefore, assume that c T di ≤ 0 for all i = 1, . . . , l (in which case, we may simply choose<br />
µi = 0, for all i). Since the set of extreme points x1, . . . xk is finite, we can simply set λp = 1<br />
if c T xp has the largest value among all possible values of c T xi, i = 1, . . . , k. This is clearly<br />
the solution to the linear programming problem. Since xp is an extreme point, we have<br />
shown that if P has a solution, it must have an extreme point solution. <br />
Corollary 5.2. Problem P has a finite solution if and only if c T di ≤ 0 for all i = 1, . . . l<br />
when d1, . . . , dl are the extreme directions of X.<br />
Proof. This is implicit in the proof of the theorem. <br />
Corollary 5.3. Problem P has alternative optimal solutions if there are at least two<br />
extreme points xp and xq so that c T xp = c T xq and so that xp is the extreme point solution<br />
to the linear programming problem.<br />
Proof. Suppose that xp is the extreme point solution to P identified in the proof of<br />
the theorem. Suppose xq is another extreme point solution with c T xp = c T xq. Then every<br />
convex combination of xp and xq is contained in X (since X is convex). Thus every x with<br />
form λxp + (1 − λ)xq and λ ∈ [0, 1] has objective function value:<br />
λc T xp + (1 − λ)c T xq = λc T xp + (1 − λ)c T xp = c T xp<br />
which is the optimal objective function value, by assumption. <br />
Exercise 48. Let X = {x ∈ R n : Ax ≤ b, x ≥ 0} and suppose that d1, . . . dl are the<br />
extreme directions of X (assuming it has any). Show that the problem:<br />
(5.6)<br />
min c T x<br />
s.t. Ax ≤ b<br />
x ≥ 0<br />
has a finite optimal solution if (and only if) c T dj ≥ 0 for k = 1, . . . , l. [Hint: Modify the<br />
proof above using the Cartheodory characterization theorem.]<br />
2. Algorithmic Characterization of Extreme Points<br />
In the previous sections we showed that if a linear programming problem has a solution,<br />
then it must have an extreme point solution. The challenge now is to identify some simple<br />
way of identifying extreme points. To accomplish this, let us now assume that we write X<br />
as:<br />
(5.7) X = {x ∈ R n : Ax = b, x ≥ 0}<br />
Our work in the previous sections shows that this is possible. Recall we can separate A into<br />
an m × m matrix B and an m × (n − m) matrix N and we have the result:<br />
(5.8) xB = B −1 b − B −1 NxN<br />
We know that B is invertible since we assumed that A had full row rank. If we assume that<br />
xN = 0, then the solution<br />
(5.9) xB = B −1 b<br />
70