1 / 25

Kernelization

Mashrur Alam Khan School of Computer Science University of Waterloo Canada. Kernelization. Outline. Background Materials A Closer Look at Some Parameterized Problems p -MAX-3-SAT p -FAST p -d-Hitting Set p -Dominating Set on Planar Graph p -Max Leaf Tree Conclusions and Future Work.

selah
Télécharger la présentation

Kernelization

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. MashrurAlam Khan School of Computer Science University of Waterloo Canada Kernelization

  2. Outline • Background Materials • A Closer Look at Some Parameterized Problems • p-MAX-3-SAT • p-FAST • p-d-Hitting Set • p-Dominating Set on Planar Graph • p-Max Leaf Tree • Conclusions and Future Work

  3. Tractability • For tractable problems, the running time of the program is a polynomial function of the input size n, say nc • For intractable problems, it is an exponential function of cn • Usually, for constant c and large n, cn> nc

  4. How to deal with Intractability • Find an improved algorithm • Unlikely, because most computer scientists believe that P ≠ NP • Or, associate a parameter with each intractable problem • Introduction to parameterized algorithm

  5. The Basic Idea Brute Force Approach O (const|x|) Input instance, x Solution Parameter, k An unique mapping from (x, k) to (x’, k’) Runtime = map(x) * const|x’| Brute Force Approach O (const|x’|) Input instance (x’, k’) Solution

  6. Kernelization • The transformation from (x, k) to (x’, k’) • Obviously, the size of |x’| must not be dependant on the size of |x| anymore • Formally, • |x’| ≤ |kernel_size(k)|, where g is a function • (x’, k’) Єinput_set <=> (x, k) Єinput_set • Also, |k’| ≤ |k|

  7. Types of Kernelization • Determined from the function of the size of kernel, kernel_size • Linear • Polynomial • Exponential (to be covered in the advanced kernelization seminar)

  8. Fixed Parameter Tractability • Runtime • O (map(x) * const|x’|) • ≤ O (map(x) * const|kernel_size(k)|) • Fixed Parameter Tractability • p-Vertex Cover • Non-FPT • p-Dominating Set, p-Clique • Beyond the scope of the presentation • We will stick to FPT, linear and polynomial kernels

  9. Outline • Background Materials • A Closer Look at Some Parameterized Problems • p-MAX-3-SAT • p-FAST • p-d-Hitting Set • p-Dominating Set on Planar Graph • p-Max Leaf Tree • Conclusions and Future Work

  10. p-MAX-3-SAT • Problem • (a1V a2V a3) ^ (a1V a2V a4) ^ … ^ (a6V a8V an) • n variables, m clauses, parameter = k • Input size, |x| = m ≤ 3n • Is there any assignment of variables which satisfies at least k of these clauses? 1 2 m

  11. p-MAX-3-SAT • Observation • Consider any arbitrary assignment • a1= 1, a2= 1, a3= 0, …, an - 1 = 1, an= 0 • Does it satisfy m/2 clauses? • If yes, move to the next slide • If no, then its complement will • a1= 0, a2= 0, a3= 1, …, an - 1 = 0, an= 1

  12. p-MAX-3-SAT • If k < m / 2 • Trivially yes • If k ≥ m / 2 • m ≤ 2k • Delete the variables that do not occur in any clause • n ≤ 6k, because each clause contain at most three new variables • We have obtained a kernel of size 6k • Linear kernel

  13. Outline • Background Materials • A Closer Look at Some Parameterized Problems • p-MAX-3-SAT • p-FAST • p-d-Hitting Set • p-Dominating Set on Planar Graph • p-Max Leaf Tree • Conclusions and Future Work

  14. p-FAST-Background • Tournament • A directed graph where each pair of vertices are connected by an arc • Feedback Arc Set • A set of arcs whose removal turns a tournament into an acyclic graph 1 2 3 4 5

  15. p-FAST

  16. Outline • Background Materials • A Closer Look at Some Parameterized Problems • p-MAX-3-SAT • p-FAST • p-d-Hitting Set • p-Dominating Set on Planar Graph • p-Max Leaf Tree • Conclusions and Future Work

  17. p-d-Hitting Set • Universe, U • {0, 1, 2, 3, 4, 5, 6, 7, 8, 9} • Family of sets with fixed cardinality, F • {{0, 1, 2}, {4, 0, 2}, {0, 6, 1}, …, {3, 8, 7}} • Cardinality of sets within F, d • 3 • Parameter, k • Is there a subset of U, U’? • U’ = {0, 1, …, 3} and |U’| ≤ k

  18. p-d-Hitting Set • Sunflower • A collection of sets with common intersection • S = {S1, S2, …, Sk} • S1Ω S2 = S2Ω S3 = S1Ω S3 = … = SiΩSj = … = Sk-1ΩSk = C • Assume we have a sunflower S in F, |S| = k + 1 • Throw away the unnecessary parts and obtain new F’ • <U, F, k> = <U, F’, k> • Kernel size will be O(kd * d!) sets (Theorem)

  19. Outline • Background Materials • A Closer Look at Some Parameterized Problems • p-MAX-3-SAT • p-FAST • p-d-Hitting Set • p-Dominating Set on Planar Graph • p-Max Leaf Tree • Conclusions and Future Work

  20. p-Dominating Set on Planar Graph • Given a planar graph G (V, E) and a parameter k, is there a dominating set of size at most k in G? • Data reduction rule (Produces linear kernel) • For any vertex, v,

  21. Outline • Background Materials • A Closer Look at Some Parameterized Problems • p-MAX-3-SAT • p-FAST • p-d-Hitting Set • p-Dominating Set on Planar Graph • p-Max Leaf Tree • Conclusions and Future Work

  22. p-Max Leaf Spanning Tree • Given a graph G (V, E) and a parameter k, find a spanning tree T in that graph with at least k leaves (if exists). • Data reduction rule (Produces linear kernel)

  23. Outline • Background Materials • A Closer Look at Some Parameterized Problems • p-MAX-3-SAT • p-FAST • p-d-Hitting Set • p-Dominating Set on Planar Graph • p-Max Leaf Tree • Conclusions and Future Work

  24. Conclusions and Future Work • We have seen • Kernelization basics • A few examples on how it works • We will see (in the advanced kernelization seminar) • Exponential Kernels • Bounds on kernel size • Lower • Upper • Recent developments on the bounds

  25. Questions? Thank You

More Related