Design and Analysis of Algorithms

CS345

Lecture 32
• NP Completeness – III
• Coping with NP-completeness

1
 NP Complete problems

Is P = NP ?

P NP

NP-complete

If any NP-complete problem is solved in polynomial time

 P = NP
 Showing a problem to be NP-complete

Let be a problem which we wish to show to be NP-complete

1. Show that ϵ NP
2. Pick a problem which is already known to be NP-complete
3. Show that

NP
𝑨
𝑿
 EXAMPLE

Showing Dominating Set to be NP-complete

4
 Dominating Set
Definition: Given an undirected graph ,
= {| or () }
In other words, is the set consisting of and all neighbors of in .
a subset is said to be an dominating set if
For each , IsIsititthe
thesmallest
smallest
dominating
dominatingset
set??
𝒃 NO.
YES

𝒂 𝒙

𝒛 𝒚 𝒕
Optimization version: compute dominating set of smallest size.
Decision version: Does there exist a dominating set of size ?

5
 Dominating Set
Definition: Given an undirected graph ,
= {| or () }
In other words, is the set consisting of and all neighbors of in .
a subset is said to be a dominating set if
For each ,

Decision version:
Does there exist a dominating set of size ?

Efficient Verifier:
Input: (, , ),
Behavior:
It checks
1. || = ?
2. for each , whether ?
This algorithm takes O() time.  Dominating Set NP 6
 VC DS

VC: Vertex Cover DS: Dominating Set

Input: a graph and Input: a graph and
Problem: Does there exist a vertex Problem: Does there exist a dominating
cover of size ? set of size ?

𝒃 𝒃

𝒂 𝒙 𝒂 𝒙

𝒛 𝒚 𝒕 𝒛 𝒚 𝒕

7
 VC DS

Observation 1: Let be vertex cover of .

is also a dominating set for provided there is no isolated vertex in .


So without loss of generality assume does not have any isolated vertex.

8
 VC DS

(, )
𝒇 (, )

Aim: Given a graph , how should transform it to graph such that

has a vertex cover of size if and only if has a dominating set of size .

9
 VC DS

VC: Instance (, ) DS: Instance (, )

𝒃 𝒃

𝒂 𝒙 𝒂 𝒙

𝒛 𝒚 𝒕 𝒛 𝒕
𝒚

Theorem :
has a vertex cover of size if and only if has a dominating set of size .

10
 VC DS

Theorem ():
If has a vertex cover of size
𝒗 𝒚 then has a dominating set of size .
𝒗 Proof:
Let be a vertex cover of of size .
𝒘 𝒚
Consider any vertex .
Case 1:
 is dominated (use Observation 1)
Case 2:
(,) and the edge is covered by ,
so either or .
 is dominated in this case as well 

11
 VC DS
Replace by .
If is already in , Theorem ():
then just remove . If has a dominating set of size ,
then has a vertex cover of size .
𝒗 Proof:
Let be a dominating set of .
𝒘 𝒚
We can assume that .
𝒗 Now consider any edge (,) .
𝒘 𝒚 Since is dominated in .
 or .
We are done.
Lemma:
If has a dominating set of size ,
then also has a dominating set
of size . 12
MORE NP-COMPLETE PROBLEMS

13
 Subgraph Isomorphism
 𝟏 𝟐

𝒂 𝒖
 𝒃 𝒗 𝟔
 𝟑
 𝒄 𝒘

𝟓 𝟒
Definition: Graph is said to be isomorphic to
if there is a bijection :  such that for any two vertices
if and only if
Problem: Given two graphs and ,
¿
does there exist a subgraph of which is isomorphic to ?

Homework: Show that subgraph isomorphism problem is NP-complete.

14
 Subset sum problem
Problem:
Given a set of integers: , ,…, ,
does there exist any subset such that

Showing that subset sum problem is in NP: easy

Showing that subset sum problem is NP-complete :
neither Homework
nor Exam problem

15
 HOW TO HANDLE
NP-COMPLETE PROBLEMS

Approximation Algorithms

16
 Approximation algorithms

• Vertex cover problem:

There is an O() time algorithm that computes a vertex cover such that

• Dominating set problem:

There is an O() time algorithm that computes a dominating set such that

• Traveling Salesman Problem:

If triangle inequality holds : approximation ratio
If triangle inequality does not hold: no constant approximation ratio is possible

17
VERTEX COVER PROBLEM

an O() time algorithm

-approximate solution

18
 The algorithm

;
While <> empty do
{ Let () be any edge present in ;
;
Remove every edge incident on ;
?
Remove every edge incident on ;
}
Return

Observation: The set is always a vertex cover of the given graph.

Homework: Prove that

19
SET COVER PROBLEM

20
 Set Cover Problem
– a set = {, ,…,}
– , ,…, , with
Optimization version: Compute least number of sets that can cover .
Decision version: Does there exist subsets that can cover all elements ?

𝒆𝟐 𝒆𝟕

𝒆𝟓
𝒆𝟏
𝒆𝟑 𝒆𝟖
𝒆𝟒 𝒆𝟔
𝒆𝟗

21
 Set Cover Problem
– a set = {, ,…,}
– , ,…, , with
Optimization version: Compute least number of sets that can cover .
Decision version: Does there exist subsets that can cover all elements ?

𝒆𝟐 𝒆𝟕

𝒆𝟓
𝒆𝟏
𝒆𝟑 𝒆𝟖
𝒆𝟒 𝒆𝟔
𝒆𝟗

22
 Set Cover Problem
– a set = {, ,…,}
– , ,…, , with
Optimization version: Compute least number of sets that can cover .
Decision version: Does there exist subsets that can cover all elements ?

𝒆𝟐 𝒆𝟕

𝒆𝟓
𝒆𝟏
𝒆𝟑 𝒆𝟖
𝒆𝟒 𝒆𝟔
𝒆𝟗

23
 Set Cover Problem
– a set = {, ,…,}
– , ,…, , with
Optimization version: Compute least number of sets that can cover .
Decision version: NP-complete Homework

Approximation algorithm for the optimization version:

;
While <> empty do
{ Pick a set such that is maximum;
Remove all elements of from ;
}
Return all subsets picked in the while loop.

Think simple.
Follow your intuition
24