15.

093 Optimization Methods

Lecture 6: Duality Theory II

c c a1 A a2 B c a5 a1 c a4 C a1 c a1 a5 D a1 a2 a4 a3

a3

Outline
Geometry of duality The dual simplex algorithm Farkas lemma Duality as a proof technique

Slide 1

The Geometry of Duality

Slide 2 min s.t. c x a i x bi , i = 1, . . . , m

max p b m pi ai = c s.t.
i=1

p0

3
3.1

Dual Simplex Algorithm

Motivation
Slide 3 In simplex method B 1 b 0 Primal optimality condition
1 c c A 0 BB

same as dual feasibility

a2

c a1 a3 a2

a3

x *

a1

Simplex is a primal algorithm: maintains primal feasibility and works towards dual feasibility Dual algorithm: maintains dual feasibility and works towards primal feasibility Slide 4 c B xB xB (1) . . . xB (m) Do not require B 1 b 0 Require c 0 (dual feasibility) Dual cost is
1 p b = c b = c BB B xB

c 1 | B 1 A 1 |

... ...

c n | B 1 An |

If B 1 b 0 then both dual feasibility and primal feasibility, and also

same cost optimality
Otherwise, change basis

3.2

An iteration

Slide 5

1. Start with basis matrix B and all reduced costs 0. 2. If B 1 b 0 optimal solution found; else, choose l s.t. xB (l) < 0.

x2

p2

1 c

. .

D b C

. . . .
B C A 1

. .
E 2 A x1 B 1/2

.
(b)

.
E 1

p1

(a)

3. Consider the lth row (pivot row) xB (l) , v1 , . . . , vn . If i vi 0 then dual

optimal cost = + and algorithm terminates.
Slide 6 4. Else, let j s.t. c j c i = min |vj | {i|vi <0} |vi |

5. Pivot element vj : Aj enters the basis and AB (l) exits.

3.3

An example
min x1 + x2 s.t. x1 + 2x2 2 x1 1 x1 , x2 0 min s.t. x1 + x2 x1 + 2x2 x3 = 2 x1 x4 = 1 x1 , x2 , x3 , x4 0 x1 0 x3 = x4 = 2 1 1 1 1 x2 1 2* 0 max s.t. 2 p1 + p2 p1 + p2 1 2 p1 1 p1 , p2 0 x4 0 0 1

Slide 7

Slide 8

x3 0 1 0

Slide 9 x1 1 x2 = x4 = 1 1 1/2 1/2 1* 3 x2 0 1 0 x3 1/2 1/2 0 x4 0 0 1

A
2

p b

x1 3/2 x2 = x1 = 1/2 1 0 0 1

x2 0 1 0

x3 1/2 1/2 0

x4 1/2 1/2 1

4
4.1

Duality as a proof method

Farkas lemma
Slide 10

Theorem:
Exactly one of the following two alternatives hold:
1. x 0 s.t. Ax = b. 2. p s.t. p A 0 and p b < 0. 4.1.1 Proof

Slide 11

If x 0 s.t. Ax = b, and if p A 0 , then p b = p Ax 0 Assume there is no x 0 s.t. Ax = b (P ) max 0 x s.t. Ax = b x 0 (D) min s.t. p b p A 0

(P) infeasible (D) either unbounded or infeasible Since p = 0 is feasible (D) unbounded p : p A 0 and p b < 0

MIT OpenCourseWare http://ocw.mit.edu

15.093J / 6.255J Optimization Methods

Fall 2009

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.