• Nem Talált Eredményt

# Minicourse on parameterized algorithms and complexity Part 1: Basic techniques

N/A
N/A
Protected

Ossza meg "Minicourse on parameterized algorithms and complexity Part 1: Basic techniques"

Copied!
42
0
0

Teljes szövegt

(1)

U N I V E R S I T Y O F B E R G E N

### Part 1: Basic techniques

Dániel Marx

(slides by Bart M. P. Jansen) November 2, 2016

(2)

### Why we are here

To create the recipes that make computers solve our problems efficiently

With a bounded number of resources (memory, time)

We measure the quality of an algorithm by the dependence of its running time on the size of the input

For an -bit input, the running time can be

Smaller functions are better, but as a general guideline:

Polynomials are good, exponential functions are bad

Unfortunately, many problems are NP-complete

We believe that for NP-complete problems, there is no algorithm that:

always gives the right answer, and whose

running time is bounded by a polynomial function of the input size

2

(3)

### Dealing with NP-complete problems

Approximation Sacrifice quality of the solution: quickly find a solution that is provably not very bad

Local search Quickly find a solution for which you cannot give any quality guarantee (but which might often be good)

Branch &

bound

Sacrifice running time guarantees: create an algorithm for which you do not know how long it will take (but which might do well on the inputs you use)

Parameterized algorithms

Sacrifice the running time: allow the running time to have an exponential factor, but ensure that the exponential dependence is not on the entire input size but just on some parameter that is hopefully small

Kernelization Quickly shrink the input by preprocessing so that afterward running an exponential-time algorithm on the shrunk instance is fast enough

3

(4)

4

(5)

### Parameterized problems

• As usual in complexity theory, we primarily study decision problems (YES/NO questions)

– OPTIMIZATION: “Find the shortest path from to ”

– DECISION: “Is there a path from to of length at most ?”

• Having an efficient algorithm for one typically gives an efficient algorithm for the other

• A parameterized problem is a decision problem where we associate an integer parameter to each instance

– The parameter measures some aspect of the instance

5

(6)

### Problem parameterizations

• PACKET DELIVERY PROBLEM

Input: A graph , a starting vertex , a set of delivery vertices, and an integer

Question: Is there a cycle in that starts and ends in , visits all vertices in , and has length at most ?

• There are many possible parameters for this problem:

– The length of the tour

– The number of delivery points

– Graph-theoretic measures of how complex is (treewidth, cliquewidth, vertex cover number)

• Parameterized complexity investigates:

Can the problem be solved efficiently, if the parameter is small?

Can the problem be solved efficiently, if the parameter is small?

6

(7)

### Fixed-parameter tractability – informally

• A parameterized problem is fixed-parameter tractable if there is an algorithm that solves size- inputs with parameter value in time for some constant and function

For each fixed , there is a polynomial-time algorithm

• VERTEX COVER:

– “Can all the edges of this -vertex graph be covered by at most vertices?”

– Solvable in time , so FPT

7

(8)

### Fixed-parameter tractability – formally

• Let be a finite alphabet used to encode inputs – ( for binary encodings)

• A parameterized problem is a set

• The set contains the tuples where the answer to the question encoded by is yes; is the parameter

• The parameterized problem is fixed-parameter tractable if there is an algorithm that, given an input ,

– decides if belongs to or not, and

– runs in time for some function and constant

8

(9)

9

(10)

### Data reduction with a guarantee

Kernelization is a method for parameterized preprocessing

Efficiently reduce an instance to an equivalent instance of size bounded by some

One of the simplest ways of obtaining FPT algorithms

Apply a brute force algorithm on the shrunk instance to get an FPT algorithm

Kernelization also allows a rigorous mathematical analysis of efficient preprocessing

10

(11)

### The VERTEX COVER problem

Input: An undirected graph and an integer Parameter:

Question: Is there a set of at most vertices in , such that each edge of has an endpoint in ?

• Such a set S is a vertex cover of

11

(12)

### Reduction rules for VERTEX COVER – (R1)

(R1) If there is an isolated vertex , delete from Reduce to the instance

12

(13)

### Reduction rules for VERTEX COVER – (R1)

(R1) If there is an isolated vertex , delete from Reduce to the instance

To ensure that a reduction rule does not change the answer, we have to prove safeness of the reduction rule

If is transformed into then we should prove that:

is a YES-instance is a YES-instance

13

(14)

### Reduction rules for VERTEX COVER – (R1)

(R1) If there is an isolated vertex , delete from Reduce to the instance

14

(15)

15

### ( �′ = ,�′ = 6 )

(R2) If there is a vertex of degree more than , then delete (and its incident edges) from and decrease the parameter by 1

Reduce to the instance

(16)

### Reduction rules for VERTEX COVER – (R3)

(R3) If the previous rules are not applicable and has more than vertices or more than edges, then conclude that we are dealing with a NO-instance

16

(17)

### Correctness of the cutoff rule

Claim. If is exhaustively reduced under (R1)-(R2) and has more than vertices or edges, then there is no size- vertex cover

Proof.

Suppose has a vertex cover

Since (R1) does not apply, every vertex of has at least one edge Since (R2) does not apply, every vertex has degree at most :

So

So if has a size- vertex cover, and

17

(18)

### Preprocessing for VERTEX COVER

• (R1)-(R3) can be exhaustively applied in polynomial time

• In polynomial time, we can reduce a VERTEX COVER instance to an instance such that:

– the two instances are equivalent: has answer YES if and only if has answer YES

– instance has at most vertices and edges

• This gives an FPT algorithm to solve an instance : – Compute reduced instance

– Solve by brute force: try all vertex subsets

For each , test if it is a vertex cover of size at most

18

(19)

### Kernelization – formally

Let be a parameterized problem and

A kernelization (or kernel) for of size is an algorithm that, given , takes time polynomial in , and outputs an instance such that:

A polynomial kernel is a kernel whose function is a polynomial

19

Theorem. A parameterized problem is fixed-

parameter tractable if and only if it is decidable and has a kernel (of arbitrary size)

Theorem. A parameterized problem is fixed-

parameter tractable if and only if it is decidable and has a kernel (of arbitrary size)

(20)

### Kernel for FEEDBACK ARC SET IN TOURNAMENTS

Input: A tournament and an integer Parameter:

Question: Is there a set of at most directed edges in , such that is acyclic?

20

(21)

### Reduction rules for FEEDBACK ARC SET

(R1) If vertex is not in any triangle, then remove

(R2) If edge is in at least distinct triangles, reverse it and decrease by one

(R3) If the previous rules are not applicable and has more than vertices, then conclude that we are dealing with a NO-instance

21

(22)

### High-level kernelization strategy

• Compare to VERTEX COVER:

– (R1) deals with elements that do not constrain the solution – (R2) deals with elements that must be in any solution

– (R3) deals with graphs that remain large after reduction

22

(23)

23

(24)

### Background

• A branching algorithm that explores a search tree of

bounded depth is one of the simplest types of FPT algorithms

• Main idea:

– Reduce problem instance to solving a bounded number of instances with parameter

• If you can solve in polynomial time using the answers to two instances and , then the problem can be solved in time

– (assuming the case is polynomial-time solvable)

• If you generate subproblems instead of 2, then the problem can be solved in time

24

(25)

7

7

88

99

1010

1111

1212

1313

1414

3

3

44

55

66

1

1

22

25

(26)

### Analysis of bounded-depth search trees

• If the parameter decreases for each recursive call, the depth of the tree is at most

• # nodes in a depth- tree with leaves is

– Usually sufficient to bound the number of leaves

• If the computation in each node takes polynomial time, total running time is

26

(27)

### VERTEX COVER revisited

Input: A graph and an integer Parameter:

Question: Is there a set of at most vertices in , such that each edge has an endpoint in ?

27

(28)

### Algorithm for VERTEX COVER

Algorithm VC(Graph , integer )

if then return NO

if has no edges then return YES

else pick an edge in and let and be its endpoints – return (VC OR (VC)

• Correct because any vertex cover must use or

• A size- vertex cover in G that uses , yields a size- vertex cover in

28

(29)

### Running time for VERTEX COVER

• Every iteration either solves the problem directly or makes two recursive calls with a decreased parameter

• The branching factor of the algorithm–and therefore of the search tree–is two

• Tree of depth with branching factor has at most leaves – Running time is

– Much better than from the kernelization algorithm

• One way to faster algorithms:

– Pick a vertex of maximum degree, recurse on and

29

(30)

### The FEEDBACK VERTEX SET problem

Input: An undirected (multi)graph and an integer Parameter:

Question: Is there a set of at most vertices in , such that each cycle contains a vertex of ?

• We allow multiple edges and self-loops

• Such a set is a feedback vertex set of

– Removing from results in an acyclic graph, a forest

30

(31)

### Branching for FEEDBACK VERTEX SET

• For VERTEX COVER, we could easily identify a set of vertices to branch on: the two endpoints of an edge

• For feedback vertex set, a solution may not contain any endpoint of an edge

– How should we branch?

• We will find a set of vertices such that any size- feedback vertex set contains a vertex of

• To find we first have to simplify the graph using reduction rules that do not change the answer

31

(32)

### Reduction rules

(R1) If there is a loop at vertex , then delete and decrease by one (R2) If there is an edge of multiplicity larger than , then reduce its multiplicity to

(R3) If there is a vertex of degree at most , then delete

(R4) If there is a vertex of degree two, then delete and add an edge between ’s neighbors

32

If (R1-R4) cannot be applied anymore, then the minimum degree is at least If (R1-R4) cannot be applied anymore,

then the minimum degree is at least

Observation. If is transformed into , then:

1. FVS of size in FVS of size in

2. Any feedback vertex set in is a feedback vertex set in when combined with the vertices deleted by (R1)

Observation. If is transformed into , then:

1. FVS of size in FVS of size in

2. Any feedback vertex set in is a feedback vertex set in when combined with the vertices deleted by (R1)

(33)

### Identifying a set to branch on

• Let be a graph whose vertices have degree three or more – Order the vertices as by decreasing degree

– Let be the largest-degree vertices

Lemma. If all vertices of have degree 3 or more, then any size- feedback vertex set of contains a vertex from

• So if there is a size- solution, it contains a vertex of – For each recurse on the instance

• Gives an algorithm with running time

– Apply the reduction rules, compute , then branch

33

(34)

### A useful claim

Claim. If is a feedback vertex set of , then

Proof. Graph is a forest – So

– Every edge not in , is incident with a vertex of

• With this claim, we can prove the degree lemma

34

(35)

### Proving the degree lemma

Lemma. If all vertices of have degree 3 or more, then any size- feedback vertex set of G contains a vertex from

– Let be a size- feedback vertex set with – By choice of we have:

, so:

– Define . Since :

By the previous

claim By the previous

claim

35

(36)

### Proving the degree lemma (II)

• The degree sum counts every edge twice:

• Combining these:

• So

• But since all vertices have degree we have:

36

(37)

### A final word on bounded-depth search trees

• The degree lemma proves the correctness of our branching strategy for FEEDBACK VERTEX SET

• When building a branching algorithm for a parameterization by the solution size:

– Find an -size set that contains a vertex of the solution – Branch in directions, trying all possibilities

– We get a search tree of depth and branching factor

• You can think of the branching process as guessing

37

(38)

38

(39)

### The SET COVER problem

Input: A set family over a universe and an integer Parameter:

Question: Is there a subfamily of at most sets, such that

• The subfamily covers the universe

• SET COVER parameterized by the universe size is FPT – Algorithm with running time

– Based on dynamic programming

39

4

4

(40)

### Dynamic programming for SET COVER

• Let

• We define a DP table for and

= min nr. of sets from needed to cover Or if impossible

• The value gives the minimum size of a set cover

– To solve the problem, compute using base cases and a recurrence

40

(41)

### Filling the dynamic programming table

= min nr. of sets from needed to cover

Base case:

if , otherwise it is

Recursive step:

Skip set , or pay for and afterwards cover

Each entry can be computed in polynomial time entries in total

41

(42)

### More on dynamic programming

• Dynamic programming is a memory-intensive algorithmic paradigm that yields FPT algorithms in various situations

– Here: dynamic programming over subsets of

– Later: dynamic programming over tree decompositions

• Research challenge:

– Determine whether the factor can be improved to for some

42

Hivatkozások

KAPCSOLÓDÓ DOKUMENTUMOK

Considering the parameterized complexity of the local search approach for the MMC problem with parameter ` denoting the neighborhood size, Theorem 3 shows that no FPT local

The algorithm described in Figure 1 solves the problem by making recursive calls to itself, or calling the algorithm of Theorem 2.4 O(n 2 ) times. In the former case, at most

By definition, the removal of all simplicial vertices from a nice graph breaks all ATs, thereby yielding an interval graph. This implies that a nice graph has a very special

Task: classify which classes A of relational structures make the problem fixed-parameter tractable parameterized by the size of the query.... Two

If exists optimal LP solution that sets x v to 1, then exists optimal vertex cover that selects v..  Remove v from G and decrease k

Theorem: [Grohe, Grüber 2007] There is a polynomial-time algorithm that finds a solution of D ISJOINT DIRECTED CYCLES with OPT/̺(OPT) cycles for some nontrivial function ̺....

By default iterative compression adds n factor to the running time.. Ex: show that for VC and FVST this factor can be reduced to O(k) (hint:

The undirected approach does not work: the pushing lemma is not true.. Pushing Lemma (for