Fast Hamiltonicity Checking via Bases of Perfect Matchings

Slides:



Advertisements
Similar presentations
Solving connectivity problems parameterized by treewidth in single exponential time Marek Cygan, Marcin Pilipczuk, Michal Pilipczuk Jesper Nederlof, Dagstuhl.
Advertisements

Weighted Matching-Algorithms, Hamiltonian Cycles and TSP
Determinant Sums (and Labeled Walks) for Undirected Hamiltonicity
Divide and Conquer. Subject Series-Parallel Digraphs Planarity testing.
Theory of Computing Lecture 18 MAS 714 Hartmut Klauck.
Reducibility Class of problems A can be reduced to the class of problems B Take any instance of problem A Show how you can construct an instance of problem.
1 Steiner Tree on graphs of small treewidth Algorithms and Networks 2014/2015 Hans L. Bodlaender Johan M. M. van Rooij.
Combinatorial Algorithms
1 Discrete Structures & Algorithms Graphs and Trees: II EECE 320.
© 2006 Pearson Addison-Wesley. All rights reserved14 A-1 Chapter 14 Graphs.
Branch and Bound Similar to backtracking in generating a search tree and looking for one or more solutions Different in that the “objective” is constrained.
CSE 421 Algorithms Richard Anderson Lecture 27 NP Completeness.
GRAPH Learning Outcomes Students should be able to:
The Theory of NP-Completeness 1. Nondeterministic algorithms A nondeterminstic algorithm consists of phase 1: guessing phase 2: checking If the checking.
Programming & Data Structures
Complexity 25-1 Complexity Andrei Bulatov Counting Problems.
Techniques for Proving NP-Completeness Show that a special case of the problem you are interested in is NP- complete. For example: The problem of finding.
Data Structures & Algorithms Graphs
CSE 421 Algorithms Richard Anderson Lecture 27 NP-Completeness and course wrap up.
Lecture 6 NP Class. P = ? NP = ? PSPACE They are central problems in computational complexity.
Instructor Neelima Gupta Table of Contents Class NP Class NPC Approximation Algorithms.
Computing Branchwidth via Efficient Triangulations and Blocks Authors: F.V. Fomin, F. Mazoit, I. Todinca Presented by: Elif Kolotoglu, ISE, Texas A&M University.
LIMITATIONS OF ALGORITHM POWER
Uib.no UNIVERSITY OF BERGEN On Sparsification for Computing Treewidth Bart M. P. Jansen Insert «Academic unit» on every page: 1 Go to the menu «Insert»
Lecture 25 NP Class. P = ? NP = ? PSPACE They are central problems in computational complexity.
Exponential time algorithms Algorithms and networks.
The Theory of NP-Completeness 1. Nondeterministic algorithms A nondeterminstic algorithm consists of phase 1: guessing phase 2: checking If the checking.
Complexity 24-1 Complexity Andrei Bulatov Counting Problems.
Kernel Bounds for Path and Cycle Problems Bart M. P. Jansen Joint work with Hans L. Bodlaender & Stefan Kratsch September 8 th 2011, Saarbrucken.
The NP class. NP-completeness Lecture2. The NP-class The NP class is a class that contains all the problems that can be decided by a Non-Deterministic.
1 GRAPH Learning Outcomes Students should be able to: Explain basic terminology of a graph Identify Euler and Hamiltonian cycle Represent graphs using.
CSE 332: NP Completeness, Part II Richard Anderson Spring 2016.
ICS 353: Design and Analysis of Algorithms NP-Complete Problems King Fahd University of Petroleum & Minerals Information & Computer Science Department.
Complexity Classes.
The NP class. NP-completeness
More NP-Complete and NP-hard Problems
More NP-complete problems
Hamilton cycles in highly symmetric graphs
Algorithms for Finding Distance-Edge-Colorings of Graphs
EMIS 8373: Integer Programming
Richard Anderson Lectures NP-Completeness
Richard Anderson Lecture 26 NP-Completeness
Lecture 2-2 NP Class.
Hans Bodlaender, Marek Cygan and Stefan Kratsch
Richard Anderson Lecture 26 NP-Completeness
Hard Problems Introduction to NP
Homomorphic Hashing for Sparse Coefficient Extraction
Paths, Trees and Flowers
Perfect Matchings in Bipartite Graphs
Chapter 5. Optimal Matchings
Lecture 5 NP Class.
Dissertation for the degree of Philosophiae Doctor (PhD)
Faster Space-Efficient Algorithms for Subset Sum
Finding Large Set Covers Faster via the Representation Method
ICS 353: Design and Analysis of Algorithms
Intro to NP Completeness
Richard Anderson Lecture 25 NP-Completeness
Exponential Time Paradigms Through the Polynomial Time Lens
Richard Anderson Lecture 28 Coping with NP-Completeness
Richard Anderson Lecture 30 NP-Completeness
A Tight Lower Bound for Counting Hamiltonian Cycles via Matrix Rank
Bart M. P. Jansen Jesper Nederlof
“(More) Consequences of Falsifying SETH
NP-Completeness Reference: Computers and Intractability: A Guide to the Theory of NP-Completeness by Garey and Johnson, W.H. Freeman and Company, 1979.
The Theory of NP-Completeness
Richard Anderson Lecture 27 Survey of NP Complete Problems
Hamiltonicity below Dirac’s condition
The Polynomial Hierarchy Enumeration Problems 7.3.3
Concepts of Computation
Treewidth meets Planarity
Presentation transcript:

Fast Hamiltonicity Checking via Bases of Perfect Matchings Wordle created using wordle.net and Hamilt: 100 NP-complete: 100 Computation: 90 Creativity: 80 Extensions: 70 Automated: 70 Surprises: 70 Structured search: 60 Techniques: 60 Paradigms: 60 Verifiers: 60 Fundamental: 60 Branching: 50 Dynamic Programming: 50 Efficiently: 50 Insights:30 Solutions:30 Tasks: 30 Omni-present: 20 Marek Cygan, Stefan Kratsch, Jesper Nederlof STOC 2013

Hamiltonicity (aka Hamiltonian cycle) Held&Karp (‘61), Bellman (‘62):                time and space (Dynamic Programming). Gurevich&Shelah (‘71):                time, poly space (Divide & Conquer). Kohn et al. (‘77), Karp:                time, poly space (Inclusion-Exclusion). Björklund (‘10):                     time, poly space (Determinant approach).

Hamiltonicity (aka Hamiltonian cycle) Traveling Salesman Held&Karp (‘61), Bellman (‘62):                time and space (Dynamic Programming). Gurevich&Shelah (‘71):                time, poly space (Divide & Conquer). Kohn et al. (‘77), Karp:                time, poly space (Inclusion-Exclusion). Björklund (‘10):                     time, poly space (Determinant approach).

Hamiltonicity (aka Hamiltonian cycle) Decompose HC into subpaths, keep track of visited vertices (     ) to make sure subpaths combine. Held&Karp (‘61), Bellman (‘62):                time and space (Dynamic Programming). Gurevich&Shelah (‘71):                time, poly space (Divide & Conquer). Kohn et al. (‘77), Karp:                time, poly space (Inclusion-Exclusion). Björklund (‘10):                     time, poly space (Determinant approach).

Hamiltonicity (aka Hamiltonian cycle) Decompose HC into subpaths, keep track of visited vertices (     ) to make sure subpaths combine. Held&Karp (‘61), Bellman (‘62):                time and space (Dynamic Programming). Gurevich&Shelah (‘71):                time, poly space (Divide & Conquer). Kohn et al. (‘77), Karp:                time, poly space (Inclusion-Exclusion). Björklund (‘10):                     time, poly space (Determinant approach). Split input graph into two subgraphs. Decompose HC into intersection with subgraphs. Keep track of.. what exactly?

Directed Bipartite graphs

Directed Bipartite graphs

Graph with separator S            

Graph with separator 2             S

Graph with separator 2 2 2             S

Graph with separator 2 2 2       S       Seems to require storing information for every matching. ->                        

Our contribution We determine the required amount of stored information via studying the rank of the ‘Matchings Connectivity matrix’: For even integer           , and perfect matchings                 of the complete graph on                   ,                                        

Our contribution We show that and use it to over GF(2),        has rank at most              by giving an explicit basis       and corresponding factorization. changing representation between can be done quick.       induces a permutation matrix. and use it to Solve Hamiltonicity on Directed Bipartite graphs in                        time and space. Solve Hamiltonicity in                                 time and space if a path decomposition of width        is given. Exclude                                        time under SETH.

k-CNF-SAT on n vars cannot be solved in time when tends to infinity. Our contribution We show that over GF(2),        has rank at most              by giving an explicit basis       and corresponding factorization. changing representation between can be done quick.       induces a permutation matrix. and use it to Solve Hamiltonicity on Directed Bipartite graphs in                        time and space. Solve Hamiltonicity in                                 time and space if a path decomposition of width        is given. Exclude                                        time under SETH. k-CNF-SAT on n vars cannot be solved in                          time when     tends to infinity.

Base cases t=4 1

Base cases t=6

Basis of H_t We define a set of pm’s         as follows: Divide base set                    into groups: 1 | 2 3 | 4 5 | . . | . t-1 | t

Basis of H_t We define a set of pm’s         as follows: Divide base set                    into groups:       are all matchings using only edges between consecutive groups. 1-1 correspondence with               -length bitstrings 1 | 2 3 | 4 5 | . . | . t-1 | t

Basis of H_t We define a set of pm’s         as follows: Divide base set                    into groups:       are all matchings using only edges between consecutive groups. 1-1 correspondence with               -length bitstrings 1 | 2 3 | 4 5 | . . | . t-1 | t 1 1

Basis of H_t We define a set of pm’s         as follows: Divide base set                    into groups:       are all matchings using only edges between consecutive groups. 1-1 correspondence with               -length bitstrings 1 | 2 3 | 4 5 | . . | . t-1 | t 1

Basis of H_t We define a set of pm’s         as follows: Divide base set                    into groups:       are all matchings using only edges between consecutive groups. 1-1 correspondence with               -length bitstrings                           HC. iff complementary bitstrings. Let         be obtained by using order     rather then                1 | 2 3 | 4 5 | . . | . t-1 | t 1

Base cases t=6

Proof idea of basis Observation: Every matching belongs to for some . Let      be obtained from     by swapping to consecutive elements. We show that for                   , the row                  is the sum of at most 3 rows from        . This relies on the base cases: All rows share all but at most three edges, which can therefore be contracted. We are left with pm’s from       .

HC in directed bipartite graphs Focus on computing parity of number of Hamiltonian cycles. Adding isolation lemma gives monte carlo algorithm.

HC in directed bipartite graphs

HC in directed bipartite graphs

HC in directed bipartite graphs 1 1 1 1 1 1 1 1 1 1 1

HC in directed bipartite graphs 1 1 1 1 1 1 1 1 1 1 1

HC in directed bipartite graphs 01000101 1 1 1 1 1 1 1 1 1 1 1

HC in directed bipartite graphs 100100010 01000101 1 1 1 1 1 1 1 1 1 1 1

HC in directed bipartite graphs 100100010 1 01000101 1 1 1 1 1 1 1 1 1 1 1

HC in directed bipartite graphs Compute the representation amounts to, for every                   compute #given pm’s that are HC with M. We compute this representation using DP over all subsets of matchings in        . Since         is a permutation matrix, we can easily compute the number of 1’s from the two vectors in               time.

Graphs with small pathwidth 2 2 2             S

Graphs with small pathwidth Again, counted #solutions modulo 2 Use dynamic programming, for every partition of the bag into degree 0,1,2 vertices store a representation of the matchings that can be established by partial solutions. Non-trivial when a new edge e is inserted: change representation to         for some     that puts e at the end.

Further results Lower bound under Strong ETH is rather technical reduction using some generic gadgets and the permutation matrix property from the basis Recently, a superset of the authors showed that rank upper bounds such as the one presented can also be used to give deterministic algorithms, and extend to weighted problems.

Conclusions Rank upper abounds on the partial solution versus partial solutions matrix are useful for designing algorithms. Can Hamiltonicity be solved in                   time? Can our tools be used to obtain faster deterministic algorithms and algorithms for TSP?

Conclusions Rank upper abounds on the partial solution versus partial solutions matrix are useful for designing algorithms. Can Hamiltonicity be solved in                   time? Can our tools be used to obtain faster deterministic algorithms and algorithms for TSP? Thanks for attending!