1 / 50

On the Complexity of K-Dimensional-Matching

On the Complexity of K-Dimensional-Matching. Elad Hazan, Muli Safra & Oded Schwartz. Maximal Matching in Bipartite Graphs. Maximal Matching in Bipartite Graphs. Easy problem: in P. 3-Dimensional Matching (3-DM). 3-Dimensional Matching (3-DM). Matching in a bounded hyper-graph.

keira
Download Presentation

On the Complexity of K-Dimensional-Matching

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. On the Complexity of K-Dimensional-Matching Elad Hazan, Muli Safra & Oded Schwartz

  2. Maximal Matching in Bipartite Graphs

  3. Maximal Matching in Bipartite Graphs Easy problem: in P

  4. 3-Dimensional Matching (3-DM)

  5. 3-Dimensional Matching (3-DM) Matching in a bounded hyper-graph Bounded Set Packing NP-hard [Karp72]

  6. Bounded variant: App. : [HS89] Inapp. : [CC03] Set-Packing: [BH92] [Hås99] 3-DM: Bounded Set-Packing Maximal Matching in a Hyper-Graph which is 3-uniform & 3-strongly-colorable

  7. K

  8. K

  9. Bounded variant: App. : [HS89] Inapp. : [Tre01] Set-Packing: [BH92] [Hås99] k-DM: Bounded Set-Packing Maximal Matching in a Hyper-Graph which is k-uniform & k-strongly-colorable Without this this is k-SP

  10. Unless P=NP, k-DM cannot be approximated to within Main Theorem: Corollary: The same holds for k-Set-Packing and Independent set in k+1-claw-free graphs Some inapproximability factors for small k-values are also obtained

  11. Gap-Problems and Inapproximability Maximization problem A Gap-A-[sno, syes]

  12. Gap-Problems and Inapproximability Maximization problem A Gap-A-[sno, syes] is NP-hard.  Approximating A better than syes/sno is NP-hard.

  13. Gap-Problems and Inapproximability Gap-k-DM-[ ] is NP-hard.  k-DM is NP-hard to approximate to within

  14. x1 + x2 + x3 = a1 mod q x7 + x4 + x2 = a2 mod q … x8 + x2 + x9 = an mod q L-q: Input: A set of linear equations mod q Objective: Find an assignment satisfying maximal number of equations App. ratio: 1/q Inapp. factor: 1/q+ [Hås97]

  15. x1 + x2 + x3 = a1 mod q x7 + x4 + x2 = a2 mod q … x8 + x2 + x9 = an mod q Thm [Hås97]: Gap-L-q-[1/q+,1-] is NP-hard. Even if each variable x occurs a constant number of times, cx = cx()

  16. x1 + x2 + x3 = a1 mod q x7 + x4 + x2 = a2 mod q … x8 + x2 + x9 = an mod q Gap-L-q ≤p Gap-k-SP Can be extended to k-DM

  17. x1 + x2 + x3 = a1 mod q x7 + x4 + x2 = a2 mod q … x8 + x2 + x9 = an mod q • Gap-L-q ≤p Gap-k-SP •   H = (V,E) • We describe hyper edges, then which vertices they include. 1st trial:

  18. 1 : x1 + x2 + x3 = 0 mod 3 A(1)=(0,1,2) 2 : x7 + x4 + x2 = 1 mod 3 A(2)=(1,0,0) x1 + x2 + x3 = a1 mod q x7 + x4 + x2 = a2 mod q … x8 + x2 + x9 = an mod q • 1st trial: • Gap-L-q ≤p Gap-k-SP • A hyper-edge for each equation and a satisfying assignment to it (q2 such assignments).

  19. 1 : x1 + x2 + x3 = 0 mod 3 A(1)=(0,1,2) 2 : x7 + x4 + x2 = 1 mod 3 A(2)=(1,0,0) x2:(1,0) x1 + x2 + x3 = a1 mod q x7 + x4 + x2 = a2 mod q … x8 + x2 + x9 = an mod q • 1st trial: • Gap-L-q ≤p Gap-k-SP • A hyper-edge for each equation and a satisfying assignment to it • A common vertex for each two contradicting edges

  20. 1 : x1 + x2 + x3 = 0 mod 3 A(1)=(0,1,2) 2 : x7 + x4 + x2 = 1 mod 3 A(2)=(1,0,0) x2:(1,0) x1 + x2 + x3 = a1 mod q x7 + x4 + x2 = a2 mod q … x8 + x2 + x9 = an mod q 1st trial: Gap-L-q ≤p Gap-k-SP Maximal matching Consistent assignment

  21. x1 + x2 + x3 = a1 mod q x7 + x4 + x2 = a2 mod q … x8 + x2 + x9 = an mod q 1st trial: Gap-L-q ≤p Gap-k-SP Maximal matching Consistent assignment Gap-L-q-[1/q+,1- ] <p Gap-k-SP-[1/q+,1- ] What is k ? k is large ! k  (cx1+cx2+cx3)q(q-1)

  22. x2=0 x2=1 x2=2 x1 + x2 + x3 = a1 mod q x7 + x4 + x2 = a2 mod q … x8 + x2 + x9 = an mod q • Gap-L-q ≤p Gap-k-SP • Saving a factor of q: • Reuse vertices • k Still depends on cx1+cx2+cx3 • which depends on 

  23. x1 + x2 + x3 = a1 mod q x7 + x4 + x2 = a2 mod q … x8 + x2 + x9 = an mod q • 2nd trial: • Gap-L-q ≤p Gap-k-SP • Allow pluralism: • A (few) contradicting edges may reside in a matching • Common vertices for only somesubsets of contradicting edges • - using a connection scheme.

  24. cx q Which contradicting edges to connect ? A Connection Scheme for x Fewer vertices: Consistency achieved using disperser-Like Properties

  25. Def:[HSS03] -Hyper-Disperser H=(V,E) V=V1 V2 … Vq E V1 × V2 × … × Vq Uindependent set (of the strong sense) i, |U\Vi| < |V| If U is large it is concentrated ! This generalizes standard dispersers

  26. Lemma [HSS03]: Existence of -Hyper-Disperser q>1,c>1 1/q2-Hyper-Disperser which is also q uniform, q strongly-colorable d regular, d strongly-edge-colorable for d=(q log q) Proof… Optimality…

  27. Def:[HSS03] -Hyper-Edge-Disperser H=(V,E) E=E1 E2 … Eq M matching i, |M\Ei| < |E| If M is large it is concentrated !

  28. Lemma [HSS03]: Existence of -Hyper-Edge-Disperser q>1,c>1 1/q2-Hyper-Edge-Disperser which is also q regular, q strongly-edge-colorable d uniform, d strongly-colorable for d=(q log q) Jump…

  29. (c=cx). •  x - a copy of • V  the vertices of all • Constructing the k-SP instance •   H =(V,E)

  30. 1 X1 X2 0 3 X3 Constructing the k-SP instance   H =(V,E) • E  for each equation  and a satisfying assignment to it – the union of three hyper-edges : x1 + x2 + x3 = 4 A()=(0,1,3) e,(0,1,2) H is 3d uniform 3d=(q log q)

  31. Completeness: • If A satisfying 1- of  • then • M covering 1- of V (hence of size |V|/k) Proof: Take all edges corresponding to the satisfying assignment. ڤ

  32. Soundness: If A satisfies at most 1/q + of  then M covers at most 4/q2 +  of V

  33. A  most popular values of each Soundness-Proof: Mmaj  Edges of M that agree with A Mmin  M \ Mmaj (Håstad)

  34. Every edge of Mmin is a minority in at least one Soundness-Proof:

  35. Soundness-Proof:

  36. Unless P=NP, k-SP cannot be approximated to within Gap-L-q-[1/q+ ,1- ] ≤p Gap-k-SP- [O(1/q),1- ] What is k ?  Gap-k-SP-[ ] is NP-hard. k=3d=(q log q) 

  37. Unless P=NP, k-SP cannot be approximated to within Conclusion Deterministic reduction This can be extended for k-DM. 4-DM, 5-DM and 6-DM cannot be approximated to within respectively.

  38. Open Problems Low-Degree: 3-DM,4-DM… TSP Steiner-Tree Sorting By Reversals

  39. Open Problems Separating k-IS from k-DM ? [HS89] [Vis96] [HSS03] [Tre01]

  40. THE END

  41. Optimality of Hyper-Disperser: 1/q2-Hyper-Disperser Regularity: d=(q log q) Restrict hyper disperser to V1,V2. A bipartite -Disperser is of degree (1/ log 1/) and   1/q. Definition…

  42. Existence of Hyper-Disperser Proof: random construction. Random permutations: ji R Sc j{2,…,q}, i[d] e[i,j] = { v[1,j], v[2, 2i(j)], …, v[q, ki(j)] } E = {e[i,j] | j{2,…,q}, i[d] } Definition…

  43. Proof – cont. Candidates: ‘bad’ (minimal) sets: U = { U | U  V, |U| = 2c/q, |UV1|=c/q}

  44. Proof – cont.

  45. Proof – cont.

  46. Gap-k-SP-[O(log k / k),1-] is NP-hard. Extending it to k-DM

  47. Use a for each location of a variable. Gap-k-DM-[O(log k / k),1-] is NP-hard.

  48. From Asymptotic to Low Degree – • How to make k as small as possible ? • Minimize d ( = 3) – by minimizing q ( = 2)(a bipartite disperser) • Avoid union of edges

  49. X2 X1 X3 From Asymptotic to Low Degree – How to make k as small as possible ? • E   equation and a satisfying assignment to it –three hyper-edges e,(0,1,2),x1 : x1 + x2 + x3 = 0 A()=(0,1,1) e,(0,1,2),x2 e,(0,1,2),x3

More Related