introduction and overview of fpt algorithmics

30
AGAPE Corsica 2009 Introduction and Introduction and overview of FPT overview of FPT algorithmics algorithmics Michael Fellows University of Newcastle, Australia

Upload: reina

Post on 21-Jan-2016

62 views

Category:

Documents


0 download

DESCRIPTION

Introduction and overview of FPT algorithmics. Michael Fellows University of Newcastle, Australia. What is FPT?. A quick overview of parameterized complexity “A two-dimensional sequel to P vs NP and all that.” “An opening chapter of multivariate complexity analysis and algorithm design.”. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

Introduction and overview of Introduction and overview of FPT algorithmicsFPT algorithmics

Michael Fellows

University of Newcastle, Australia

Page 2: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

• A quick overview of parameterized complexity

• “A two-dimensional sequel to P vs NP and all that.”

• “An opening chapter of multivariate complexity analysis and algorithm design.”

What is FPT?

Page 3: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

The “classical” P vs NP framework is one-dimensional

n = input size

poly(n) 2 poly(n)vs

“good”

P

positive toolkit of how to design P-time algorithms

“bad”

NP, etc.

negative toolkit of NP-hardness, etc.

Unfortunately, almost everything turns out to be NP-hard.

Page 4: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

The parameterized framework is two-dimensional

n = input sizek = a relevant secondary measure

f(k)nc n g(k)vs

“good”

FPT

“bad”

W-hard, etc.

Complexity frameworks are driven by contrasting function classes.

Page 5: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

Frameworks in pictures

The classical P vs NP framework

k

nc

nIntrinsic Combinatorial explosion: Most problems are NP-hard or worse.

The parameterized framework

FPT

Try to confine the explosion to the parameter.

Page 6: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

The main parameterized hierarchy:current

P lin(k) poly(k) FPT

M[1] W[1] M[2]

W[2] ... W[SAT] W[P]

... XP

P

The best kind of FPT is P

kernelization classes

A simple view: (problems of interest) + (reductions) =

empirical complexity classes.

Page 7: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

W[1] is the natural two-dimensional analog of NP

Our premier guides to intractability are various forms of the HALTING PROBLEM.

•HP I

in A program P (Turing machine Mp)? Will it ever halt?

Undecidable

•HP II

in A Turing machine M nondeterministic.? Can M halt in |M | steps?

•HP III in A nondeterministic Turing machine M (unlimited nondeterminism & alphabet size), k.? Can M halt in < k steps?

Trivially (by def.) complete for NP

Complete for W[1]

Page 8: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

W[1] is the natural two-dimensional analog of NP

Our premier guides to intractability are various forms of the HALTING PROBLEM.

•HP I

in A program P (Turing machine Mp)? Will it ever halt?

Undecidable

•HP II

in A Turing machine M nondeterministic.? Can M halt in |M | steps?

•HP III in A nondeterministic Turing machine M (unlimited nondeterminism & alphabet size), k.? Can M halt in < k steps?

Trivially (by def.) complete for NP

Complete for W[1]

Page 9: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

Downey-Fellows: Parameterized Complexity, Springer, 1999

Flum-Grohe: Parameterized Complexity Theory, Springer, 2006.

Niedermeier: Invitation to Fixed-Parameter Algorithms, Oxford University Press, 2006.

Page 10: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

Rodney G. Downey, Michael R. Fellows, and Michael A. Langston Foreword by the Guest Editors Falk Hüffner, Rolf Niedermeier, and Sebastian Wernicke Techniques for Practical Fixed-Parameter Algorithms Michael A. Langston, Andy D. Perkins, Arnold M. Saxton, Jon A. Scharff, and Brynn H. Voy Innovative Computational Methods for Transcriptomic Data Analysis: A Case Study in the Use of FPT for Practical Algorithm Design and Implementation Jianer Chen and Jie Meng On Parameterized Intractability: Hardness and Completeness Dániel Marx Parameterized Complexity and Approximation Algorithms Jens Gramm, Arfst Nickelsen, and Till Tantau Fixed-Parameter Algorithms in Phylogenetics Leizhen Cai Parameterized Complexity of Cardinality Constrained Optimization Problems Christian Sloper and Jan Arne Telle An Overview of Techniques for Designing Parameterized Algorithms

Book review    William Gasarch and Keung Ma Kin

Invitation to Fixed-Parameter Algorithms • Parameterized Complexity Theory • Parameterized Algorithmics: Theory, Practice and Prospects

volume 51, number 1 January 2008

the Computer journal

Page 11: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

volume 51, number 3 May 2008

Hans L. Bodlaender and Arie M. C. A. Koster Combinatorial Optimization on Graphs of Bounded TreewidthLiming Cai, Xiuzhen Huang, Chunmei Liu, Frances Rosamond, and Yinglei Song Parameterized Complexity and Biopolymer Sequence Comparison Erik D. Demaine and MohammadTaghi Hajiaghayi The Bidimensionality Theory and Its Algorithmic Applications Georg Gottlob and Stefan Szeider Fixed-Parameter Algorithms For Artificial Intelligence, Constraint Satisfaction and Database Problems

Petr Hlineny, Sang-il Oum, Detlef Seese, and Georg GottlobWidth Parameters Beyond Tree-width and their Applications Gregory Gutin and Anders Yeo Some Parameterized Problems On Digraphs

Panos Giannopoulos, Christian Knauer, and Sue Whitesides Parameterized Complexity of Geometric Problems Iris van Rooij and Todd Wareham Parameterized Complexity in Cognitive Modeling: Foundations, Applications and Opportunities

the Computer journal

Page 12: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

Page 13: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

• How to design efficient algorithms

The positive toolkit of FPT algorithm techniques.

• How to analyze complexity and recognize intractability.

The negative toolkit of M[1] and W[1] hardness.

Two complementary mathematical toolkits:

Page 14: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

Page 15: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

for k = number of cities by “try all permutations”

implies k! n FPT

Shows that a parameterized problem can be “trivially FPT”

But it is still interesting to look for “better FPT”

Such as 2kn

Of course TSP is also in FPT

Page 16: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

(1) The “f(k) race”: The race to findBetter and better (slower growing) f(k)

(2) The “kernelization race”: The race to find

Smaller and smaller P-time kernelizations

When a parameterized problem is shown FPT, two races begin

Page 17: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

SET SPLITTING:In: Family F 2X of subsets of a base set X.

Parameter: k

Question: Does there exist X’ X that that splits at least k sets in F ? Where X’ splits S F if there exists a S, a X – X’ and there exists a’ S, a’ X’.

Dehne et al (2003): O(72k n3)Dehne et al (2004): O*(8k) “Crown reduction”Lokshtanov and Sloper (2005): O*(2.7k)Chen, Liu (2008):O*(2k) “Randomized disposal”

Another example: SET SPLITTING

f(k) r

ace

Page 18: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

Page 19: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

Page 20: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

UNDIRECTED FEELBACK VERTEX SET:In: (G, k)

Parameter: k

Question: Is it possible to delete at most k vertices from G to get a graph G’ that is acyclic?

• Known to be FPT for many years, but no poly(k) kernelization

• Burrage et al. (2006) O(k11) P-time kernelization• Bodlaender et al. (2007) O(k3) kernel• Thomasse (2009) O(k2) kernel

Example of a “kernelization race”

Page 21: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

(1) VERTEX COVER

In: (G, k)

Parameter: k

Question: Are there k vertices that cover all the edges?

(2) CLIQUE

In: (G, k) Parameter: k

Question: Does G have a k clique?

(3) DOMINATING SET

In: (G, k) Parameter: k

Question: Are there k vertices that cover all the vertices of G?

3 simple graph problems that drove the fieldNP-complete

but P for any fixed k

NP-complete

but P for any fixed k

NP-complete

but P for any fixed k

Page 22: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

(1) VERTEX COVER is linear-time for any fixed k by “bounded search tree” approach

(2) Both k-CLIQUE and k-DOM SET seem to require something more like brute force: “try all k-subsets” naively O(n k+1)

What was found

Page 23: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

Page 24: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

Page 25: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

Page 26: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

Page 27: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

Page 28: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

Page 29: Introduction and overview of FPT algorithmics

AGAPECorsica 2009

Page 30: Introduction and overview of FPT algorithmics

AGAPECorsica 2009