polyhedral optimization lecture 3 – part 3 m. pawan kumar [email protected] slides available...

25
Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar [email protected] Slides available online http://cvn.ecp.fr/personnel/pawan/

Upload: eleanor-gilmore

Post on 22-Dec-2015

218 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

Polyhedral OptimizationLecture 3 – Part 3

M. Pawan Kumar

[email protected]

Slides available online http://cvn.ecp.fr/personnel/pawan/

Page 2: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

Solving Linear Programs

s.t. A x ≤ b

maxx cTxOptimization

Feasibility asks if there exists an x such that

cTx ≥ K

A x ≤ b

Optimization via binary search on K

Feasible solution

For a given K

Page 3: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

Feasibility via Ellipsoid Method

Feasible region of LP

Page 4: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

Feasibility via Ellipsoid Method

Ellipsoid containing feasible region of LP

Page 5: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

Feasibility via Ellipsoid Method

Centroid of ellipsoid

Page 6: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

Feasibility via Ellipsoid Method

Separating hyperplane for centroid

Page 7: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

Feasibility via Ellipsoid Method

Smallest ellipsoid containing “truncated” ellipsoid

Page 8: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

Feasibility via Ellipsoid Method

Centroid of ellipsoid

Page 9: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

Feasibility via Ellipsoid Method

Separating hyperplane for centroid

Page 10: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

Feasibility via Ellipsoid Method

Smallest ellipsoid containing “truncated” ellipsoid

Page 11: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

Feasibility via Ellipsoid Method

Centroid of ellipsoid

Page 12: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

Feasibility via Ellipsoid Method

Terminate when feasible solution is found

Page 13: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

• Separating hyperplane in polynomial time– Check each of the ‘m’ LP constraints in O(n) time

• New ellipsoid in polynomial time– Shor (1971), Nemirovsky and Yudin (1972)

• Polynomial iterations (Khachiyan 1979, 1980)– Volume of ellipsoid reduces exponentially

• Only requires a separation oracle– Constraint matrix A can be very large

Ellipsoid Method

Page 14: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

• Separation implies easy optimization

• What about the reverse?

• Matroid polytopes admit greedy optimization

• Do they allow easy separation?

• Why are we even interested in this?

Optimization vs. Separation

Page 15: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

• Polar Polyhedron

• Using Optimization for Separation

• Poly-Time Equivalence

Outline

Page 16: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

Polar Polyhedron

Polyhedron P = {x: Ax ≤ b}

Polar Polyhedron P* = {y: for all x P, ∈ xTy ≤ 1}

Assume 0 is in the interior of P

(P*)* = P Proof?

b > 0

No “loss of generality” as P can be translated

Page 17: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

P is a subset of (P*)*

If x P, then for all ∈ y P* we have ∈ xTy ≤ 1

(P*)* = {z: for all y P*, ∈ zTy ≤ 1}

Therefore, x (P*)*∈

Page 18: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

(P*)* is a subset of P

Let there be an x P∉

There must exist a separating hyperplane

cTx > d cTz ≤ d, for all z P ∈

Since 0 interior of P, d > 0 ∈

Without loss of generality, d = 1

Page 19: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

(P*)* is a subset of P

Let there be an x P∉

There must exist a separating hyperplane

cTx > 1 cTz ≤ 1, for all z P ∈

c P* ∈

x (P*)*∉

Why?

Why? Hence proved

Page 20: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

• Polar Polyhedron

• Using Optimization for Separation

• Poly-Time Equivalence

Outline

Page 21: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

Optimization Problem over P

Polyhedron P = {x: Ax ≤ b}

max cTx

x P ∈

Page 22: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

Separation Problem over P*

Polar Polyhedron P* = {y: for all x P, ∈ xTy ≤ 1}

Given y, return ‘YES’ if y P* ∈

Otherwise, return separating hyperplane

Page 23: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

Using Optimization for Separation

Set c = y

max cTx

x P ∈

C* =

If C* ≤ 1, then return ‘YES’

If C* > 1, then return x*

Optimal solution x*

Page 24: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

• Polar Polyhedron

• Using Optimization for Separation

• Poly-Time Equivalence

Outline

Page 25: Polyhedral Optimization Lecture 3 – Part 3 M. Pawan Kumar pawan.kumar@ecp.fr Slides available online

Poly-Time Equivalence

OptimizationonP

SeparationonP*

Polarity

OptimizationonP*

Ellipsoidmethod

Separationon

(P*)* = P

Polarity

Ellipsoidmethod