1 combinatorial dominance analysis keywords: combinatorial optimization (co) approximation...

Post on 21-Dec-2015

239 Views

Category:

Documents

5 Downloads

Preview:

Click to see full reader

TRANSCRIPT

1

Combinatorial Dominance Analysis

Keywords:Combinatorial Optimization (CO) Approximation Algorithms (AA)Approximation Ratio (a.r)Combinatorial Dominance (CD)Domination number/ratio (domn, domr)DOM-good approximationDOM-easy problem

by: Yochai Twitto

2

Overview

Background On approximations and approximation

ratio.

Combinatorial Dominance What is it ? Definitions & Notations.

Problem: maximum Cutmaximum Cut Summary

3

Overview

Background On approximations and approximation

ratio.

Combinatorial Dominance What is it ? Definitions & Notations.

Problem: maximum Cutmaximum Cut Summary

4

Background NP complexity class.

AA and quality of approximations.

The classical approximation ratio analysis.

Example: Approximation for TSP.

5

NP

If P ≠ NP, then finding the optimum of NP-hard problem is difficult.

If P = NP, P would encompass the NP and NP-Complete areas.

6

Approximations

So we are satisfied with an approximate solution.

Question: How can we measure

the solution quality ?Solutions

quality line

OPT

Infeasible

Near optimal

7

Solution Quality

Most of the time, naturally derived from the problem definition.

If not, it should be given as external information.

8

The classical Approximation Ratio(For maximization problem)

Assume 0 ≤ β ≤ 1. A.r. ≥ β if

the solution quality is greater than β·OPT

Solutions quality line

OPT

Infeasible

Near optimal

½OPT

9

Example:

The Traveling Salesman Problem

Given a weighted complete graph G, find the optimal tour.

We will assume the graph is metric.

We will see: The MST approximation. MST approximation ratio

analysis.

10

MST Approximation for TSP

Find a minimum spanning tree for G.

DFS the tree. Make shortcuts.

11

MST Approx. ratio analysis Observation:

If you remove an edge from a tour then you get a spanning tree!

This means that Tour cost more

than a minimum spanning tree.

12

MST Approx. ratio analysis Thus, DFSing the MST

is of cost No more than twice

MST cost. I.e. no more than twice

OPT.

After shortcuts we get a tour with cost at most twice the optimum

Since the graph is metric.

13

Overview

Background On approximations and approximation

ratio.

Combinatorial Dominance What is it ? Definitions & Notations.

Problem: maximum Cutmaximum Cut Summary

14

Combinatorial Dominance

What is a “combinatorial dominance guarantee” ?

Why do we need such guarantees ?

Example: the min partition min partition problemproblem.

Definitions and notations.

15

What is a

“combinatorial dominance guarantee” ?

A letter of reference: “She is half as good as I am, but I am the best

in the world…” “she finished first in my class of 75 students…”

The former is akin to an approximation ratio.

The latter to combinatorial dominance guarantee.

16

What is a

“combinatorial dominance guarantee” ? (cont.)

We saw that MST provides a 2-factor approximation.

We can ask: Is the returned

solution guaranteed to be always in the top O(n) best solutions ?

Solutions quality line

OPT

Infeasible

Near optimaltop

O(n)

17

Why do we need that ?

Let us take another lookLet us take another look at the MST approximation for TSP.

All other edges of weight 1+ε

(not shown)

18

Why do we need that ?

The spanning tree here is a star. DFS + Shortcuts yields

OPT = 6 + 4ε ≈ 6

MST tour size: 10

In general:

OPT: (n-2)(1+ε) + 2

MST: 2(n-2) + 2

OPT

MST tour

19

Why do we need that ? But this is the worst possible tour! Such kind of analysis is called blackball

analysis.

Blackball instance

20

Corollary

The approximation ratio analysis gives us only a partial insight of the performance of the algorithm.

Dominance analysis makes the picture fuller.

21

Simple example of dominance analysis

The minimum partition problem.

Greedy-type algorithm.

Combinatorial dominance analysis of the algorithm.

22

Example:

The minimum partition problem

Given is a set of n numbers V = { a1, a2, …, an}

Find a bipartition (X,Y ) of the indices such that

is minimal.

Yi iXi i aaYXf ),(

23

Greedy-type algorithm

Without loss of generality assumea1 ≥ a2 ≥ … ≥ an .

Initiate X = { }, Y = { } . For j = 1, …, n

Add j to X if , Otherwise add j to Y .

Yi iXi i aa

24

Combinatorial dominance analysis of the greedy-type algorithm

Observation:Any solution produced by the alg. satisfies .

Assume (X ’,Y ’) is any solution for min partition for {a2, a3, …, an}.

Now, add a1 to Y ’ if ,

Otherwise add a1 to X ’.

1),( aYXf

Yi iXi i aa

25

Combinatorial dominance analysis of the greedy-type algorithm (cont.)

Obtained solution: (X ’’,Y ’’). (X ’’, Y ’’) is a solution of the

original problem. We have Conclusion:

The solution provided by the algorithm dominates at least 2n-1

solutions.

),()'',''( YXfYXf

26

Definitions & Notations

Domination number: domn Domination ratio: domr

DOM-good approximation DOM-easy problem

27

Domination Number: domn Let P be a CO problem. Let A be an approximation for P .

For an instance I of P, the domination number domn(I, A) of A on I is the number of feasible solutions of I that are not better than the solution found by A.

28

domn (example)

STSP on 5 vertices. There exist 12 tours

If A returns a tour of length 7 then domn(I, A) = 8

4, 5, 5, 6, 7, 9, 9, 11, 11, 12, 14, 14

(tours lengths)

29

Domination Number: domn Let P be a CO problem. Let A be an approximation for P .

For any size n of P, the domination number domn(P, n, A) of an approximation A for P is the minimumminimum of domn(I, A) over all instances I of P of size n.

30

Domination Ratio: domr Let P be a CO problem. Let A be an approximation for P . Denote by sol(sol(I I )) the number of all

feasible solutions of I.

For any size n of P, the domination ratio domn(P, n, A) of an approximation A for P is the minimumminimum of domn(I, A) / sol(I ) taken over all instances I of P of size n.

31

DOM-good approximation

A is a DOM-good approximation algorithm for P, if It is a polynomial time complexity alg. There exists a polynomial p(n) in the

size of P, such that The domination ratio of A is at least

1/p(n) for any size n of P.

32

DOM-easy problem

A CO problem is a DOM-easy problem if it admits a DOM-good approximation.

Problems not having this property are DOM-hard.

Corollary:Minimum Partition is DOM-easy. Furthermore, p(n) is a constant.

33

Overview

Background On approximations and approximation

ratio.

Combinatorial Dominance What is it ? Definitions & Notations.

Problem: Maximum CutMaximum Cut Summary

34

Maximum Cut

The problem. Simple greedy algorithm. Combinatorial dominance of the

algorithm.

We’ll see…

Maximum Cut is DOM-easy.

35

Problem: Maximum Cut Input: weighted complete graph G=(V, E,

w) Find a bipartition (X, Y) of V maximizing

the sum

Denote n = |V|. Let W be the sum of weights of all edges.

)()(

XYYXEeew

36

Problem: Maximum Cut

Denote the average weight of a cut by

Notice that . Next:

We’ll see a simple algorithm which produces solutions that are always better than .

We’ll show it is a DOM-good approximation for maxCut.

2/WW

W

W

37

Algorithm: greedy maxCut Algrorithm:

Initiate X = {}, Y = {} For each j = 1…n

Add vj to X or Y so as to maximize its marginal value.

Theorem: The above algorithm is a 2-factor

approximation for maxCut. Moreover, it produces a cut of weight at least

.W

38

CD analysis

We will show that the number of cuts of weight at most is at least a polynomial part of all cuts Call them “bad” cuts

Note that this is a general analysis technique. Can be applied to another

algs./problems

W

39

CD analysis

A k-cut is a cut (X, Y) for which |X| = k.

A fixed edge crosses k-cuts.

Hence the average weight of a k-cut is

1

22k

n

W

k

n

k

n

Wk

1

22

40

CD analysis

Let bk be the number of bad k-cuts. i.e. k-cuts of weight less than .

Then

k

k

WW

k

n

bk

n

W

41

CD analysis

Solving for bk we get

nk

n

nn

knk

k

nbk /1

)1(

)(41

nn

k 22

42

CD analysis

Hence the number of bad cuts in G is at least

(by DeMoivre-Laplace theorem)

n

nnkc

nk

n

n2

1122/

cconstant somefor

43

CD analysis

Thus, G has more than bad cuts.

Corollary:Maximum Cut is DOM-easy.

nc n2

44

Overview

Background On approximations and approximation

ratio.

Combinatorial Dominance What is it ? Definitions & Notations.

Problem: maximum Cutmaximum Cut Summary

45

Summary

Solutions quality line

OPT

Infeasible

Near optimal

½OPT

Solutions quality line

OPT

Infeasible

Near optimaltop

O(n)

46

Summary

MST tour

OPT

47

Summary

Domination number: domn Domination ratio: domr

DOM-good approximation DOM-easy problem

48

Summary

Domn(MST, TSP) = 1

Minimum Partition is DOM-easy. Maximum Cut is DOM-easy.

Clique is DOM-hard unless P=NP.

blackba

ll

49

Combinatorial Dominance Analysis

top related