kernelization algorithms for graph and other structure modification problems

105
Kernelization algorithms for graph and other structure modification problems Anthony PEREZ Supervisors: St ´ ephane BESSY and Christophe PAUL (AlGCo Team) November 14

Upload: anthony-perez

Post on 05-Dec-2014

571 views

Category:

Technology


1 download

DESCRIPTION

Thesis defense on November 14th, 2011, in Montpellier.Jury:Stéphane Bessy, Bruno Durand, Frédéric Havet, Rolf Niedermeier, Christophe Paul & Ioan Todinca.

TRANSCRIPT

Page 1: Kernelization algorithms for graph and other structure modification problems

Kernelization algorithms for graph and otherstructure modification problems

Anthony PEREZ

Supervisors: Stephane BESSY and Christophe PAUL

(AlGCo Team)

November 14

Page 2: Kernelization algorithms for graph and other structure modification problems

INTRODUCTION

(Graph) Modification problems

Input: A graph (or another structure) and a (graph) property G.Output: A minimum number of modification of the graph in order tosatisfy G.

modification: adding edges, deleting edges, deleting vertices, ...

2 / 42

Page 3: Kernelization algorithms for graph and other structure modification problems

INTRODUCTION

(Graph) Modification problems

Input: A graph (or another structure) and a (graph) property G.Output: A minimum number of modification of the graph in order tosatisfy G.

modification: adding edges, deleting edges, deleting vertices, ...

2 / 42

Page 4: Kernelization algorithms for graph and other structure modification problems

INTRODUCTION

(Graph) Modification problems

CLUSTER EDITING

Input: A graph G = (V ,E).Output: A set F ⊆ (V × V ) of minimum size such that the graphH = (V ,E M F ) is a disjoint union of cliques.

3 / 42

Page 5: Kernelization algorithms for graph and other structure modification problems

INTRODUCTION

(Graph) Modification problems

CLUSTER EDITING

Input: A graph G = (V ,E).Output: A set F ⊆ (V × V ) of minimum size such that the graphH = (V ,E M F ) is a disjoint union of cliques.

3 / 42

Page 6: Kernelization algorithms for graph and other structure modification problems

INTRODUCTION

(Graph) Modification problems

Cover a broad range of NP-Hard problems:

VERTEX COVER

FEEDBACK VERTEX SET

More general: F -MINOR DELETION

EDGE-MULTICUT

4 / 42

Page 7: Kernelization algorithms for graph and other structure modification problems

INTRODUCTION

(Graph) Modification problems

Find applications in various domains:

bioinformaticsmachine learningrelational databasesimage processing

4 / 42

Page 8: Kernelization algorithms for graph and other structure modification problems

INTRODUCTION

Different approaches

Most modification problems are NP-hard.How to solve them efficiently?

Approximation algorithmsExact exponential algorithmsPreprocessing steps (heuristics)

5 / 42

Page 9: Kernelization algorithms for graph and other structure modification problems

INTRODUCTION

Different approaches

Most modification problems are NP-hard.How to solve them efficiently?

Approximation algorithmsExact exponential algorithmsPreprocessing steps (heuristics)

5 / 42

Page 10: Kernelization algorithms for graph and other structure modification problems

INTRODUCTION

Different approaches

Most modification problems are NP-hard.How to solve them efficiently?

Approximation algorithmsExact exponential algorithmsPreprocessing steps (heuristics)

How to measure the efficiency of heuristics?

5 / 42

Page 11: Kernelization algorithms for graph and other structure modification problems

INTRODUCTION

Different approaches

Most modification problems are NP-hard.How to solve them efficiently?

Approximation algorithmsExact exponential algorithmsPreprocessing steps (heuristics)

Exploit the fact that the number of modifications needed should besmall compared to the instance size n.

5 / 42

Page 12: Kernelization algorithms for graph and other structure modification problems

Outline of the talk

1 Parameterized complexity

Part I. Graph Modification Problems

2 Branches and generic reduction rules

3 PROPER INTERVAL COMPLETION

Part II. Different modification problems

4 Considered problems

5 FEEDBACK ARC SET IN TOURNAMENTS

Page 13: Kernelization algorithms for graph and other structure modification problems

PARAMETERIZED COMPLEXITY

Parameterized problem

G-MODIFICATION

Input: A graph G = (V ,E), k ∈ N.Parameter: k .Output: A set F ⊆ (V × V ) of size at most k such that the graphH = (V ,E M F ) belongs to G.

Idea. Measure the complexity of a problem with respect tosome parameter k .

7 / 42

Page 14: Kernelization algorithms for graph and other structure modification problems

PARAMETERIZED COMPLEXITY

Parameterized problem

G-MODIFICATION

Input: A graph G = (V ,E), k ∈ N.Parameter: k .Output: A set F ⊆ (V × V ) of size at most k such that the graphH = (V ,E M F ) belongs to G.

Parameterized algorithmA problem parameterized by some k ∈ N admits a parameterizedalgorithm if it can be solved in time f (k) · nO(1).

7 / 42

Page 15: Kernelization algorithms for graph and other structure modification problems

PARAMETERIZED COMPLEXITY

Kernels

Given an instance (I, k) of a parameterized problem L,a kernelization algorithm:

runs in time Poly(|I|+ k)

and outputs an instance (I′, k ′) such that:(i) (I, k) ∈ YES ⇔ (I′, k ′) ∈ YES(ii) |I′| 6 h(k) and k ′ 6 k

(I, k) (I ′, k ′)

|I ′| 6 h(k)k ′ 6 k

Poly(|I|+ k)

Do all parameterized problems admit polynomial kernels?

8 / 42

Page 16: Kernelization algorithms for graph and other structure modification problems

PARAMETERIZED COMPLEXITY

Kernels

Given an instance (I, k) of a parameterized problem L,a kernelization algorithm:

runs in time Poly(|I|+ k)

and outputs an instance (I′, k ′) such that:(i) (I, k) ∈ YES ⇔ (I′, k ′) ∈ YES(ii) |I′| 6 h(k) and k ′ 6 k

Theorem (Folklore)Parameterized algorithm⇔ Kernelization algorithm

Do all parameterized problems admit polynomial kernels?

8 / 42

Page 17: Kernelization algorithms for graph and other structure modification problems

PARAMETERIZED COMPLEXITY

Kernels

Given an instance (I, k) of a parameterized problem L,a kernelization algorithm:

runs in time Poly(|I|+ k)

and outputs an instance (I′, k ′) such that:(i) (I, k) ∈ YES ⇔ (I′, k ′) ∈ YES(ii) |I′| 6 h(k) and k ′ 6 k

Size: super-polynomial

Do all parameterized problems admit polynomial kernels?

8 / 42

Page 18: Kernelization algorithms for graph and other structure modification problems

PARAMETERIZED COMPLEXITY

Kernels

Given an instance (I, k) of a parameterized problem L,a kernelization algorithm:

runs in time Poly(|I|+ k)

and outputs an instance (I′, k ′) such that:(i) (I, k) ∈ YES ⇔ (I′, k ′) ∈ YES(ii) |I′| 6 h(k) and k ′ 6 k

Size: super-polynomial

Do all parameterized problems admit polynomial kernels?

8 / 42

Page 19: Kernelization algorithms for graph and other structure modification problems

PARAMETERIZED COMPLEXITY

Lower bounds for kernels

There exist some parameterized problems that do not admit polynomialkernels. (under a complexity-theoretic assumption)

(i) Or-composition [Bodlaender et al., 2008 - Fortnow and Santhanam, 2008]

(ii) Polynomial time and parameter transformations[Bodlaender et al., 2009]

(iii) Cross-composition [Bodlaender et al., 2011]

9 / 42

Page 20: Kernelization algorithms for graph and other structure modification problems

Graph modification problems

2 Branches and generic reduction rules

3 PROPER INTERVAL COMPLETION

G-MODIFICATION

Input: A graph G = (V ,E), k ∈ N.

Parameter: k .

Output: A set F ⊆ (V × V ) of size at most k s.t. the graph H = (V ,E M F ) belongs to G.

Page 21: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Generic reduction rules

Connected component.

If G is hereditary and closed under disjoint union, remove anyconnected component C that belongs to G.

11 / 42

Page 22: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Generic reduction rules

Sunflower.

Consider a finite forbidden induced subgraph of G (obstruction).For any pair e ⊆ (V ×V ) that belongs to a set of k + 1 obstructionspairwise intersecting exactly in e, transform G into (V ,E M {e}).

12 / 42

Page 23: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Generic reduction rules

Sunflower.

Consider a finite forbidden induced subgraph of G (obstruction).For any pair e ⊆ (V ×V ) that belongs to a set of k + 1 obstructionspairwise intersecting exactly in e, transform G into (V ,E M {e}).

12 / 42

Page 24: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Generic reduction rulesCritical clique.

Assume G is hereditary and closed under true twin addition.For any critical clique T with |T | > k + 1, remove |T | − (k + 1)arbitrary vertices from T .

v

u

13 / 42

Page 25: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Generic reduction rulesCritical clique.

Assume G is hereditary and closed under true twin addition.For any critical clique T with |T | > k + 1, remove |T | − (k + 1)arbitrary vertices from T .

v

u

13 / 42

Page 26: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Generic reduction rulesCritical clique.

Assume G is hereditary and closed under true twin addition.For any critical clique T with |T | > k + 1, remove |T | − (k + 1)arbitrary vertices from T .

Lemma [Bessy, Paul and P., 2010]

There always exists an optimal editionthat preserves the critical cliques.

k = 1

k = 1

13 / 42

Page 27: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Branches: a natural idea

Reduce set of vertices that induce a graph belonging to G.The Connected Component rule is a Branch reduction rule.

Context: can be used on problems where G admits a so-calledadjacency decomposition.

Branch: set of vertices B ⊆ Vsuch that:

(i) G[B] ∈ G and,(ii) B is connected properly

to the rest of the graph.

B

G[B] ∈ G

G \ B

14 / 42

Page 28: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Branches: a natural idea

Reduce set of vertices that induce a graph belonging to G.The Connected Component rule is a Branch reduction rule.

Context: can be used on problems where G admits a so-calledadjacency decomposition.

Branch: set of vertices B ⊆ Vsuch that:

(i) G[B] ∈ G and,(ii) B is connected properly

to the rest of the graph.

B

G[B] ∈ G

G \ B

14 / 42

Page 29: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Branches: a natural idea

Reduce set of vertices that induce a graph belonging to G.The Connected Component rule is a Branch reduction rule.

Context: can be used on problems where G admits a so-calledadjacency decomposition.

Branch: set of vertices B ⊆ Vsuch that:

(i) G[B] ∈ G and,(ii) B is connected properly

to the rest of the graph.

B

G[B] ∈ G

G \ B

14 / 42

Page 30: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Branches: a natural idea

Reduce set of vertices that induce a graph belonging to G.The Connected Component rule is a Branch reduction rule.

Context: can be used on problems where G admits a so-calledadjacency decomposition.

Branch: set of vertices B ⊆ Vsuch that:

(i) G[B] ∈ G and,(ii) B is connected properly

to the rest of the graph.

B

G[B] ∈ G

G \ B

14 / 42

Page 31: Kernelization algorithms for graph and other structure modification problems

Outline

2 Branches and generic reduction rulesGeneric reduction rulesBranches

3 PROPER INTERVAL COMPLETION

Definition and known resultsBranchesReducing the branches

Page 32: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

DefinitionPROPER INTERVAL COMPLETION

Input: A graph G = (V ,E), k ∈ N.Parameter: k .Output: A set F ⊆ (V × V ) \ E of size at most k such thatH = (V ,E ∪ F ) is a proper interval graph.

16 / 42

Page 33: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

DefinitionPROPER INTERVAL COMPLETION

Input: A graph G = (V ,E), k ∈ N.Parameter: k .Output: A set F ⊆ (V × V ) \ E of size at most k such thatH = (V ,E ∪ F ) is a proper interval graph.

16 / 42

Page 34: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

DefinitionPROPER INTERVAL COMPLETION

Input: A graph G = (V ,E), k ∈ N.Parameter: k .Output: A set F ⊆ (V × V ) \ E of size at most k such thatH = (V ,E ∪ F ) is a proper interval graph.

16 / 42

Page 35: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

DefinitionPROPER INTERVAL COMPLETION

Input: A graph G = (V ,E), k ∈ N.Parameter: k .Output: A set F ⊆ (V × V ) \ E of size at most k such thatH = (V ,E ∪ F ) is a proper interval graph.

16 / 42

Page 36: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

DefinitionPROPER INTERVAL COMPLETION

Input: A graph G = (V ,E), k ∈ N.Parameter: k .Output: A set F ⊆ (V × V ) \ E of size at most k such thatH = (V ,E ∪ F ) is a proper interval graph.

NP-Complete [Golumbic et al., 1994]

FPT : O(24km) (motivated by applications in genomic research)[Kaplan, Shamir and Tarjan, 1994]

Polynomial kernel?

16 / 42

Page 37: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

DefinitionPROPER INTERVAL COMPLETION

Input: A graph G = (V ,E), k ∈ N.Parameter: k .Output: A set F ⊆ (V × V ) \ E of size at most k such thatH = (V ,E ∪ F ) is a proper interval graph.

Theorem [Bessy and P., 2011]

The PROPER INTERVAL COMPLETION problem admits a kernel withO(k4) vertices.

16 / 42

Page 38: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Some useful results

A graph is a proper interval graph if and only if:

it does not contain any of the following graphs as an inducedsubgraph.

claw p-cycle (p ≥ 4)3-sun net

[Wegner, 1967]

17 / 42

Page 39: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Some useful results

A graph is a proper interval graph if and only if:

its vertices admit an ordering v1 . . . vn such that:

vivj ∈ E i < j ⇒ vivl , vlvj ∈ E , i < l < j

[Looges and Olartu, 1993]

17 / 42

Page 40: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Generic reduction rules

Remarks. Proper interval graphs are hereditary and:

(i) closed under disjoint union:the Connected Component rule can be applied.

(ii) do not admit any claw or C4 as an induced subgraph:the Sunflower rule can be applied.

(iii) closed under true twin addition:the Critical Clique rule can be applied.

What about branches?

18 / 42

Page 41: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Generic reduction rules

Remarks. Proper interval graphs are hereditary and:

(i) closed under disjoint union:the Connected Component rule can be applied.

(ii) do not admit any claw or C4 as an induced subgraph:the Sunflower rule can be applied.

(iii) closed under true twin addition:the Critical Clique rule can be applied.

What about branches?

18 / 42

Page 42: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Generic reduction rules

Remarks. Proper interval graphs are hereditary and:

(i) closed under disjoint union:the Connected Component rule can be applied.

(ii) do not admit any claw or C4 as an induced subgraph:the Sunflower rule can be applied.

(iii) closed under true twin addition:the Critical Clique rule can be applied.

What about branches?

18 / 42

Page 43: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Generic reduction rules

Remarks. Proper interval graphs are hereditary and:

(i) closed under disjoint union:the Connected Component rule can be applied.

(ii) do not admit any claw or C4 as an induced subgraph:the Sunflower rule can be applied.

(iii) closed under true twin addition:the Critical Clique rule can be applied.

What about branches?

18 / 42

Page 44: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Adjacency decomposition

2 3 4 5 6 7 8 9(b)

(a)

9

2

1

4

3

5

6

7

8

7895

4321

6

1

Branches can be used on PROPER INTERVAL COMPLETION.

19 / 42

Page 45: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Adjacency decomposition

2 3 4 5 6 7 8 9(b)

(a)

9

2

1

4

3

5

6

7

8

7895

4321

6

1

Branches can be used on PROPER INTERVAL COMPLETION.

19 / 42

Page 46: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Adjacency decomposition

2 3 4 5 6 7 8 9(b)

(a)

9

2

1

4

3

5

6

7

8

7895

4321

6

1

Branches can be used on PROPER INTERVAL COMPLETION.

19 / 42

Page 47: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

How to define a branch?

Consider the structure of a solution.Look at unaffected vertices.

20 / 42

Page 48: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Branches

B1 BR B2

L b1 R C

B

bl

A subset B of V is a branch if:

(i) G[B] is a connected PIG with umbrella ordering σB = b1, . . . ,b|B|,(ii) The vertex set V \ B can be partitioned into sets L,R and C with:

no edges between B and Cevery vertex in L (resp. R) has a neighbor in BNB(L) ⊂ NB[b1] = {b1, . . . ,bl′}NB(R) ⊂ NB[b|B|] = {bl , . . . ,b|B|}NL(bi+1) ⊆ NL(bi) for every 1 ≤ i < l ′ and NR(bi) ⊆ NR(bi+1) forevery l ≤ i < |B|

21 / 42

Page 49: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Branches

B1 BR B2

L b1 b|B| R C

B

A subset B of V is a branch if:

(i) G[B] is a connected PIG with umbrella ordering σB = b1, . . . ,b|B|,(ii) The vertex set V \ B can be partitioned into sets L,R and C with:

no edges between B and Cevery vertex in L (resp. R) has a neighbor in BNB(L) ⊂ NB[b1] = {b1, . . . ,bl′}NB(R) ⊂ NB[b|B|] = {bl , . . . ,b|B|}NL(bi+1) ⊆ NL(bi) for every 1 ≤ i < l ′ and NR(bi) ⊆ NR(bi+1) forevery l ≤ i < |B|

21 / 42

Page 50: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Branches

B1 BR B2

L b1 b|B| R C

B

A subset B of V is a branch if:

(i) G[B] is a connected PIG with umbrella ordering σB = b1, . . . ,b|B|,(ii) The vertex set V \ B can be partitioned into sets L,R and C with:

no edges between B and Cevery vertex in L (resp. R) has a neighbor in BNB(L) ⊂ NB[b1] = {b1, . . . ,bl′}NB(R) ⊂ NB[b|B|] = {bl , . . . ,b|B|}NL(bi+1) ⊆ NL(bi) for every 1 ≤ i < l ′ and NR(bi) ⊆ NR(bi+1) forevery l ≤ i < |B|

21 / 42

Page 51: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Branches

B1 BR B2

L b1 b|B| R C

B

blbl ′

A subset B of V is a branch if:

(i) G[B] is a connected PIG with umbrella ordering σB = b1, . . . ,b|B|,(ii) The vertex set V \ B can be partitioned into sets L,R and C with:

no edges between B and Cevery vertex in L (resp. R) has a neighbor in BNB(L) ⊂ NB[b1] = {b1, . . . ,bl′}NB(R) ⊂ NB[b|B|] = {bl , . . . ,b|B|}NL(bi+1) ⊆ NL(bi) for every 1 ≤ i < l ′ and NR(bi) ⊆ NR(bi+1) forevery l ≤ i < |B|

21 / 42

Page 52: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Branches

B1 BR B2

L b1 b|B| R C

B

bl ′ bl

A subset B of V is a branch if:

(i) G[B] is a connected PIG with umbrella ordering σB = b1, . . . ,b|B|,(ii) The vertex set V \ B can be partitioned into sets L,R and C with:

no edges between B and Cevery vertex in L (resp. R) has a neighbor in BNB(L) ⊂ NB[b1] = {b1, . . . ,bl′}NB(R) ⊂ NB[b|B|] = {bl , . . . ,b|B|}NL(bi+1) ⊆ NL(bi) for every 1 ≤ i < l ′ and NR(bi) ⊆ NR(bi+1) forevery l ≤ i < |B|

21 / 42

Page 53: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Branches

B1 BR B2

L b|B| R C

B

b1 bl ′ bl

A subset B of V is a branch if:

(i) G[B] is a connected PIG with umbrella ordering σB = b1, . . . ,b|B|,(ii) The vertex set V \ B can be partitioned into sets L,R and C with:

no edges between B and Cevery vertex in L (resp. R) has a neighbor in BNB(L) ⊂ NB[b1] = {b1, . . . ,bl′}NB(R) ⊂ NB[b|B|] = {bl , . . . ,b|B|}NL(bi+1) ⊆ NL(bi) for every 1 ≤ i < l ′ and NR(bi) ⊆ NR(bi+1) forevery l ≤ i < |B|

21 / 42

Page 54: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Branches

B1 BR B2

L b|B| R C

B

b1 bl ′ bl

A subset B of V is a branch if:

(i) G[B] is a connected PIG with umbrella ordering σB = b1, . . . ,b|B|,(ii) The vertex set V \ B can be partitioned into sets L,R and C with:

no edges between B and Cevery vertex in L (resp. R) has a neighbor in BNB(L) ⊂ NB[b1] = {b1, . . . ,bl′}NB(R) ⊂ NB[b|B|] = {bl , . . . ,b|B|}NL(bi+1) ⊆ NL(bi) for every 1 ≤ i < l ′ and NR(bi) ⊆ NR(bi+1) forevery l ≤ i < |B|

21 / 42

Page 55: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Branches

B1 BR B2

L b1 R C

B

bl

If L = ∅ (or R = ∅), B is a 1-branch, otherwise B is a 2-branchIf B is a clique, we call B a K-join

22 / 42

Page 56: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Branches

B1 BR B2

L b1 R C

B

bl

If L = ∅ (or R = ∅), B is a 1-branch, otherwise B is a 2-branchIf B is a clique, we call B a K-join

22 / 42

Page 57: Kernelization algorithms for graph and other structure modification problems

Outline

2 Branches and generic reduction rulesGeneric reduction rulesBranches

3 PROPER INTERVAL COMPLETION

Definition and known resultsBranchesReducing the branches

Page 58: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Reducing the K -joins

Cannot be done directly.

x y z t

Assuming the graph is reduced by the generic rules, we can removeO(k3) vertices from any K -join to obtain a clean K -join.

24 / 42

Page 59: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Reducing the K -joins

Cannot be done directly.

A clean K -join does not intersect any claw or C4.

Assuming the graph is reduced by the generic rules, we can removeO(k3) vertices from any K -join to obtain a clean K -join.

24 / 42

Page 60: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Reducing the K -joins

Cannot be done directly.

A clean K -join does not intersect any claw or C4.

Assuming the graph is reduced by the generic rules, we can removeO(k3) vertices from any K -join to obtain a clean K -join.

24 / 42

Page 61: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Reducing the clean K -joins

Let B be a clean K -join of size at least 2k + 2. Let Bf be the k + 1 firstvertices of B, Bl be its k + 1 last vertices and M = B \ (Bf ∪ Bl).Remove the set of vertices M from G.

Bl (k + 1 vertices)MBf (k + 1 vertices)

25 / 42

Page 62: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Reducing the clean K -joins

Let B be a clean K -join of size at least 2k + 2. Let Bf be the k + 1 firstvertices of B, Bl be its k + 1 last vertices and M = B \ (Bf ∪ Bl).Remove the set of vertices M from G.

Can be carried out in polynomial time!

25 / 42

Page 63: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Reducing the branches

In polynomial time, the 1- and 2-branches can be reduced to O(k3)vertices.

B1

Remove

B

BR

2k + 1 vertices

R G \ (B ∪ R)

26 / 42

Page 64: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Reducing the branches

In polynomial time, the 1- and 2-branches can be reduced to O(k3)vertices.

B1

Remove

B

BR

2k + 1 vertices

R G \ (B ∪ R)

Remove

B

2k + 1 vertices

BR

2k + 1 vertices

B1

LB2

RB′2B′1

26 / 42

Page 65: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Main result

Theorem [Bessy and P., 2011]

The PROPER INTERVAL COMPLETION problem admits a kernel withO(k4) vertices.

1-branch 1-branch2-branchK -join K -join K -join

27 / 42

Page 66: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Main result

Theorem [Bessy and P., 2011]

The PROPER INTERVAL COMPLETION problem admits a kernel withO(k4) vertices.

O(k3) O(k3) O(k3) O(k3) O(k3)O(k3)

1-branch 1-branch2-branchK -join K -join K -join

27 / 42

Page 67: Kernelization algorithms for graph and other structure modification problems

BRANCHES AND GENERIC REDUCTION RULES PROPER INTERVAL COMPLETION

Main result

Theorem [Bessy and P., 2011]

The PROPER INTERVAL COMPLETION problem admits a kernel withO(k4) vertices.

Related result [Bessy, Paul and P., 2010]

The CLOSEST 3-LEAF POWER problem admits a kernel with O(k3)vertices.

27 / 42

Page 68: Kernelization algorithms for graph and other structure modification problems

Different modification problems

4 Considered problems

5 FEEDBACK ARC SET IN TOURNAMENTS

Π-EDITION

Input: A dense set R of p-sized relations defined over an universe V , an integer k ∈ N.

Parameter: k .

Output: A set F ⊆ R of size at most k whose modification satisfies Π.

Page 69: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

FEEDBACK ARC SET IN TOURNAMENTS (FAST)

Input: A tournament T = (V ,A) and an integer k ∈ N.Parameter: k .Output: A set at most k arcs whose reversal results in an acyclictournament.

23 4 21

3

1 4

29 / 42

Page 70: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

FEEDBACK ARC SET IN TOURNAMENTS (FAST)

Input: A tournament T = (V ,A) and an integer k ∈ N.Parameter: k .Output: A set at most k arcs whose reversal results in an acyclictournament.

NP-Complete [Charbit et al., 2007]

Admits constant-factor approximation algorithms [Kenyon-Mathieu andSchudy, 2007]

29 / 42

Page 71: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

DENSE ROOTED TRIPLET INCONSISTENCY (RTI)

Input: A set of leaves V and a dense collection R of rooted binarytrees over three leaves of V .Parameter: k .Output: A set of at most k triplets whose modification leads to acollection admitting a consistent rooted binary tree defined over V .

ab bd dc cca b da

t1 t2 t3 t4

R := {ab|c, cd |b, ab|d , ac|d}R := {t1, t2, t3, t4}

30 / 42

Page 72: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

DENSE ROOTED TRIPLET INCONSISTENCY (RTI)

Input: A set of leaves V and a dense collection R of rooted binarytrees over three leaves of V .Parameter: k .Output: A set of at most k triplets whose modification leads to acollection admitting a consistent rooted binary tree defined over V .

a ab bd d dc cca b

a b c d

t1 t2 t3 t4

R := {ab|c, cd |b, ab|d , ac|d}

T is not consistent withR

R := {t1, t2, t3, t4}

30 / 42

Page 73: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

DENSE ROOTED TRIPLET INCONSISTENCY (RTI)

Input: A set of leaves V and a dense collection R of rooted binarytrees over three leaves of V .Parameter: k .Output: A set of at most k triplets whose modification leads to acollection admitting a consistent rooted binary tree defined over V .

a cb bd d ac dca b

a b c d

t1 t2 t3 t4

R := {ab|c, cd |b, ab|d , cd |a}

T is consistent withR

R := {t1, t2, t3, t4}

30 / 42

Page 74: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

DENSE ROOTED TRIPLET INCONSISTENCY (RTI)

Input: A set of leaves V and a dense collection R of rooted binarytrees over three leaves of V .Parameter: k .Output: A set of at most k triplets whose modification leads to acollection admitting a consistent rooted binary tree defined over V .

NP-Complete [Barky et al., 2010]

Does not admit a constant-factor approximation algorithm yet

30 / 42

Page 75: Kernelization algorithms for graph and other structure modification problems

Outline

4 Considered problemsFEEDBACK ARC SET IN TOURNAMENTS

DENSE ROOTED TRIPLET INCONSISTENCY

Conflict Packing

5 FEEDBACK ARC SET IN TOURNAMENTS

Reduction rulesConflict Packing

Page 76: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Consistency

FAST (folklore)The following properties are equivalent:

(i) T is acyclic(ii) T does not contain any directed triangle

32 / 42

Page 77: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Consistency

RTI [Guillemot and Mnich, 2010]

The following properties are equivalent:(i) R is consistent(ii) R does not contain any conflict on four leaves

Conflict. Set of vertices C ⊆ V that does not admit a consistent rooted binary tree.

32 / 42

Page 78: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Parameterized complexity

FAST RTIFPT O∗(2

√k log k ) a FPT O∗(2k1/3 log k ) b

Kernel with O(k2) verticesa Kernel with O(k2) verticesb

Linear vertex-kernel c No such result known before.

a[Alon et al., 2009]b[Guillemot and Mnich, 2010]c[Bessy et al., 2009]

The linear vertex-kernel for FAST described by [Bessy et al., 2009]uses a constant-factor approximation algorithm.Their reduction rules can be adapted to RTI.But no constant-factor approximation!

33 / 42

Page 79: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Parameterized complexity

FAST RTIFPT O∗(2

√k log k ) a FPT O∗(2k1/3 log k ) b

Kernel with O(k2) verticesa Kernel with O(k2) verticesb

Linear vertex-kernel c No such result known before.

a[Alon et al., 2009]b[Guillemot and Mnich, 2010]c[Bessy et al., 2009]

The linear vertex-kernel for FAST described by [Bessy et al., 2009]uses a constant-factor approximation algorithm.Their reduction rules can be adapted to RTI.But no constant-factor approximation!

33 / 42

Page 80: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Parameterized complexity

FAST RTIFPT O∗(2

√k log k ) a FPT O∗(2k1/3 log k ) b

Kernel with O(k2) verticesa Kernel with O(k2) verticesb

Linear vertex-kernel c No such result known before.

a[Alon et al., 2009]b[Guillemot and Mnich, 2010]c[Bessy et al., 2009]

The linear vertex-kernel for FAST described by [Bessy et al., 2009]uses a constant-factor approximation algorithm.Their reduction rules can be adapted to RTI.But no constant-factor approximation!

33 / 42

Page 81: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Parameterized complexity

FAST RTIFPT O∗(2

√k log k ) a FPT O∗(2k1/3 log k ) b

Kernel with O(k2) verticesa Kernel with O(k2) verticesb

Linear vertex-kernel c No such result known before.

a[Alon et al., 2009]b[Guillemot and Mnich, 2010]c[Bessy et al., 2009]

The linear vertex-kernel for FAST described by [Bessy et al., 2009]uses a constant-factor approximation algorithm.Their reduction rules can be adapted to RTI.But no constant-factor approximation!

33 / 42

Page 82: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Parameterized complexity

FAST RTIFPT O∗(2

√k log k ) a FPT O∗(2k1/3 log k ) b

Kernel with O(k2) verticesa Kernel with O(k2) verticesb

Linear vertex-kernel c No such result known before.

a[Alon et al., 2009]b[Guillemot and Mnich, 2010]c[Bessy et al., 2009]

The linear vertex-kernel for FAST described by [Bessy et al., 2009]uses a constant-factor approximation algorithm.Their reduction rules can be adapted to RTI.But no constant-factor approximation!

33 / 42

Page 83: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Conflict Packing

[Paul, P. and Thomasse, 2011]

works on problems characterized by some finite conflicts.maximal collection of p-uplets disjoint conflits C.provides a lower bound on the number of modification required.implies that the instance induced by V \ V (C) is consistent.

34 / 42

Page 84: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Reduction rules

Remove any vertex that is not part of any directed triangle. a.acan be carried out in polynomial time.

35 / 42

Page 85: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Reduction rules

Safe partition

Assume V (T ) is ordered under some ordering σ, and let P be apartition of σ into intervals.

Vl

AI := {uv ∈ A | ∃ i u, v ∈ Vi}

V1 V2

AO := A \ AI

B is the set of backward arcs of AO (arcs vivj with i > j).

35 / 42

Page 86: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Reduction rules

Safe partition

Assume V (T ) is ordered under some ordering σ, and let P be apartition of σ into intervals.

Vl

AI := {uv ∈ A | ∃ i u, v ∈ Vi}

V1 V2

AO := A \ AI

B is the set of backward arcs of AO (arcs vivj with i > j).

35 / 42

Page 87: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Reduction rules

Safe partition

P is safe if there exist |B| arc-disjoint conflicts using arcs of AOonly.

35 / 42

Page 88: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Safe Partition Reduction Rule

[Bessy et al., 2009]

Let P be a safe partition of an ordered tournament T = (V ,A, σ).Reverse every arc of B and decrease k accordingly.

Use constant-factor approximation algorithm.Use Conflict Packing.

36 / 42

Page 89: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Safe Partition Reduction Rule

[Bessy et al., 2009]

Let P be a safe partition of an ordered tournament T = (V ,A, σ).Reverse every arc of B and decrease k accordingly.

Use constant-factor approximation algorithm.Use Conflict Packing.

36 / 42

Page 90: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Safe Partition Reduction Rule

[Bessy et al., 2009]

Let P be a safe partition of an ordered tournament T = (V ,A, σ).Reverse every arc of B and decrease k accordingly.

Main questionHow to compute a safe partition in polynomial time?

Use constant-factor approximation algorithm.Use Conflict Packing.

36 / 42

Page 91: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Safe Partition Reduction Rule

[Bessy et al., 2009]

Let P be a safe partition of an ordered tournament T = (V ,A, σ).Reverse every arc of B and decrease k accordingly.

Main questionHow to compute a safe partition in polynomial time?

Use constant-factor approximation algorithm.Use Conflict Packing.

36 / 42

Page 92: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Safe Partition Reduction Rule

[Bessy et al., 2009]

Let P be a safe partition of an ordered tournament T = (V ,A, σ).Reverse every arc of B and decrease k accordingly.

Main questionHow to compute a safe partition in polynomial time?

Use constant-factor approximation algorithm.Use Conflict Packing.

36 / 42

Page 93: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Conflict Packing

A conflict packing of a tournament is a maximal collection ofarc-disjoint directed triangles.

37 / 42

Page 94: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Conflict Packing

A conflict packing of a tournament is a maximal collection ofarc-disjoint directed triangles.

Can be computed greedily (i.e. in polynomial time).Let C be a conflict packing. If T = (V ,A) is a positive instance then|V (C)| 6 3k .

37 / 42

Page 95: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Conflict Packing

A conflict packing of a tournament is a maximal collection ofarc-disjoint directed triangles.

Conflict Packing Lemma [Paul, P. and Thomasse, 2011]

Let T = (V ,A) be a tournament. There exists an ordering of T whosebackward arcs uv are such that u, v ∈ V (C).

37 / 42

Page 96: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Conflict Packing

A conflict packing of a tournament is a maximal collection ofarc-disjoint directed triangles.

Lemma [Paul, P. and Thomasse, 2011]

Let T = (V ,A) be a tournament such that |V | > 4k . There exists a safepartition that can be computed in polynomial time.

proof

37 / 42

Page 97: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Conflict Packing

A conflict packing of a tournament is a maximal collection ofarc-disjoint directed triangles.

Corollary [Paul, P. and Thomasse, 2011]

FEEDBACK ARC SET IN TOURNAMENTS admits a kernel with at most 4kvertices.

37 / 42

Page 98: Kernelization algorithms for graph and other structure modification problems

CONSIDERED PROBLEMS FEEDBACK ARC SET IN TOURNAMENTS

Application to the RTI problem

Remove vertices that do not belong to any conflictSafe Partition reduction ruleConflict Packing allows to find a Safe Partition

Theorem [Paul, P. and Thomasse, 2011]

DENSE ROOTED TRIPLET INCONSISTENCY admits a kernel with at most5k vertices.

38 / 42

Page 99: Kernelization algorithms for graph and other structure modification problems

Conclusion

6 Our results

7 Open problems

Page 100: Kernelization algorithms for graph and other structure modification problems

OUR RESULTS OPEN PROBLEMS

Main results

Polynomial kernelsFirst polynomial kernels:

(i) CLOSEST 3-LEAF POWER(ii) PROPER INTERVAL COMPLETION(iii) COGRAPH EDGE-EDITION

Improved polynomial kernels:

(i) FEEDBACK ARC SET IN TOURNAMENTS(ii) DENSE ROOTED TRIPLET INCONSISTENCY(iii) DENSE BETWEENNESS and DENSE CIRCULAR ORDERING

joint works with: S. Bessy, F. Fomin, S. Gaspers, S. Guillemot, F. Havet, C. Paul,S. Saurabh and S. Thomasse.

40 / 42

Page 101: Kernelization algorithms for graph and other structure modification problems

OUR RESULTS OPEN PROBLEMS

Main results

Lower bounds on kernelization:(i) For any l > 7, the Pl -FREE EDGE-DELETION problem

does not admit a polynomial kernel.(ii) For any l > 4, the Cl -FREE EDGE-DELETION problem

does not admit a polynomial kernel.

joint work with: S. Guillemot, F. Havet and C. Paul.

40 / 42

Page 102: Kernelization algorithms for graph and other structure modification problems

OUR RESULTS OPEN PROBLEMS

Open problems

Do the FEEDBACK VERTEX SET IN TOURNAMENTS and CLUSTER

VERTEX DELETION problems admit linear vertex-kernels?

Characterize lower bounds for modification problems. details

Can we use branches on other problems?(e.g. CHORDAL DELETION)Can we use Conflict Packing on other problems?(e.g. (weakly)-fragile constraint modification problems)

41 / 42

Page 103: Kernelization algorithms for graph and other structure modification problems

OUR RESULTS OPEN PROBLEMS

Open problems

Do the FEEDBACK VERTEX SET IN TOURNAMENTS and CLUSTER

VERTEX DELETION problems admit linear vertex-kernels?

Characterize lower bounds for modification problems. details

Can we use branches on other problems?(e.g. CHORDAL DELETION)Can we use Conflict Packing on other problems?(e.g. (weakly)-fragile constraint modification problems)

41 / 42

Page 104: Kernelization algorithms for graph and other structure modification problems

OUR RESULTS OPEN PROBLEMS

Open problems

Do the FEEDBACK VERTEX SET IN TOURNAMENTS and CLUSTER

VERTEX DELETION problems admit linear vertex-kernels?

Characterize lower bounds for modification problems. details

Can we use branches on other problems?(e.g. CHORDAL DELETION)Can we use Conflict Packing on other problems?(e.g. (weakly)-fragile constraint modification problems)

41 / 42

Page 105: Kernelization algorithms for graph and other structure modification problems

Merci de votre attention !