Improved Ramsey-type results in comparability graphs D aniel Kor - - PowerPoint PPT Presentation

improved ramsey type results in comparability graphs
SMART_READER_LITE
LIVE PREVIEW

Improved Ramsey-type results in comparability graphs D aniel Kor - - PowerPoint PPT Presentation

Improved Ramsey-type results in comparability graphs D aniel Kor andi EPFL May 14, 2019 joint work with Istv an Tomon Partially ordered sets Partially ordered sets Partially ordered set A poset is a set X with a transitive,


slide-1
SLIDE 1

Improved Ramsey-type results in comparability graphs

D´ aniel Kor´ andi

EPFL

May 14, 2019 joint work with Istv´ an Tomon

slide-2
SLIDE 2

Partially ordered sets

slide-3
SLIDE 3

Partially ordered sets

Partially ordered set

A poset is a set X with a transitive, antisymmetric relation <.

slide-4
SLIDE 4

Partially ordered sets

Partially ordered set

A poset is a set X with a transitive, antisymmetric relation <. ◮ t-chain: x1 < x2 < · · · < xt

slide-5
SLIDE 5

Partially ordered sets

Partially ordered set

A poset is a set X with a transitive, antisymmetric relation <. ◮ t-chain: x1 < x2 < · · · < xt ◮ r-antichain: x1, . . . , xr such that xi < xj for every i, j

slide-6
SLIDE 6

Partially ordered sets

Partially ordered set

A poset is a set X with a transitive, antisymmetric relation <. ◮ t-chain: x1 < x2 < · · · < xt ◮ r-antichain: x1, . . . , xr such that xi < xj for every i, j

Fact

Every N-element poset contains a chain of length t or an antichain

  • f size N/t (for every t).
slide-7
SLIDE 7

Partially ordered sets

Partially ordered set

A poset is a set X with a transitive, antisymmetric relation <. ◮ t-chain: x1 < x2 < · · · < xt ◮ r-antichain: x1, . . . , xr such that xi < xj for every i, j

Fact

Every N-element poset contains a chain of length t or an antichain

  • f size N/t (for every t).

“Dilworth”: If X has no t-chain, then it can be partitioned into t antichains.

slide-8
SLIDE 8

Convex sets in the plane

slide-9
SLIDE 9

Convex sets in the plane

slide-10
SLIDE 10

Convex sets in the plane

Question

Given n convex sets in the plane, how big is the largest disjoint or pairwise intersecting subfamily?

slide-11
SLIDE 11

Convex sets in the plane

Question

Given n convex sets in the plane, how big is the largest disjoint or pairwise intersecting subfamily?

slide-12
SLIDE 12

Convex sets in the plane

Question

Given n convex sets in the plane, how big is the largest disjoint or pairwise intersecting subfamily?

slide-13
SLIDE 13

Convex sets in the plane

Question

Given n convex sets in the plane, how big is the largest disjoint or pairwise intersecting subfamily?

Theorem (Larman-Matouˇ sek-Pach-T¨

  • csik, 1994)

At least n1/5.

slide-14
SLIDE 14

Convex sets in the plane

A, B disjoint, B is “below” A A B

slide-15
SLIDE 15

Convex sets in the plane

A, B disjoint, B is “below” A A B

slide-16
SLIDE 16

Convex sets in the plane

A, B disjoint, B is “below” A A B

slide-17
SLIDE 17

Convex sets in the plane

A, B disjoint, B is “below” A A B

slide-18
SLIDE 18

Convex sets in the plane

A, B disjoint, B is “below” A A B A B

slide-19
SLIDE 19

Convex sets in the plane

A, B disjoint, B is “below” A A B A B A B

slide-20
SLIDE 20

Convex sets in the plane

A, B disjoint, B is “below” A A B A B A B A B

slide-21
SLIDE 21

Convex sets in the plane

A, B disjoint, B is “below” A A B A B A B A B A <1 B A <2 B A <3 B A <4 B

slide-22
SLIDE 22

Convex sets in the plane

A, B disjoint, B is “below” A A B A B A B A B A <1 B A <2 B A <3 B A <4 B <1, <2, <3, <4 are partial orders

slide-23
SLIDE 23

Convex sets in the plane

A, B disjoint, B is “below” A A B A B A B A B A <1 B A <2 B A <3 B A <4 B <1, <2, <3, <4 are partial orders Sets A, B are disjoint iff they are comparable in any of <1, . . . , <4.

slide-24
SLIDE 24

n convex sets in the plane

Sets A, B are disjoint iff they are comparable in any of <1, . . . , <4.

slide-25
SLIDE 25

n convex sets in the plane

Sets A, B are disjoint iff they are comparable in any of <1, . . . , <4.

Fact

Every N-element poset contains a t-chain or an (N/t)-antichain.

slide-26
SLIDE 26

n convex sets in the plane

Sets A, B are disjoint iff they are comparable in any of <1, . . . , <4.

Fact

Every N-element poset contains a t-chain or an (N/t)-antichain. <1 n1/5-chain

slide-27
SLIDE 27

n convex sets in the plane

Sets A, B are disjoint iff they are comparable in any of <1, . . . , <4.

Fact

Every N-element poset contains a t-chain or an (N/t)-antichain. <1 n1/5-chain → disjoint family

slide-28
SLIDE 28

n convex sets in the plane

Sets A, B are disjoint iff they are comparable in any of <1, . . . , <4.

Fact

Every N-element poset contains a t-chain or an (N/t)-antichain. <1 n1/5-chain → disjoint family Otherwise: n4/5-antichain S1.

slide-29
SLIDE 29

n convex sets in the plane

Sets A, B are disjoint iff they are comparable in any of <1, . . . , <4.

Fact

Every N-element poset contains a t-chain or an (N/t)-antichain. <1 n1/5-chain → disjoint family Otherwise: n4/5-antichain S1. <2 n1/5-chain in S1

slide-30
SLIDE 30

n convex sets in the plane

Sets A, B are disjoint iff they are comparable in any of <1, . . . , <4.

Fact

Every N-element poset contains a t-chain or an (N/t)-antichain. <1 n1/5-chain → disjoint family Otherwise: n4/5-antichain S1. <2 n1/5-chain in S1 → disjoint family

slide-31
SLIDE 31

n convex sets in the plane

Sets A, B are disjoint iff they are comparable in any of <1, . . . , <4.

Fact

Every N-element poset contains a t-chain or an (N/t)-antichain. <1 n1/5-chain → disjoint family Otherwise: n4/5-antichain S1. <2 n1/5-chain in S1 → disjoint family Otherwise: n3/5-antichain S2 ⊆ S1.

slide-32
SLIDE 32

n convex sets in the plane

Sets A, B are disjoint iff they are comparable in any of <1, . . . , <4.

Fact

Every N-element poset contains a t-chain or an (N/t)-antichain. <1 n1/5-chain → disjoint family Otherwise: n4/5-antichain S1. <2 n1/5-chain in S1 → disjoint family Otherwise: n3/5-antichain S2 ⊆ S1. <3 n1/5-chain in S2 → disjoint family Otherwise: n2/5-antichain S3 ⊆ S2.

slide-33
SLIDE 33

n convex sets in the plane

Sets A, B are disjoint iff they are comparable in any of <1, . . . , <4.

Fact

Every N-element poset contains a t-chain or an (N/t)-antichain. <1 n1/5-chain → disjoint family Otherwise: n4/5-antichain S1. <2 n1/5-chain in S1 → disjoint family Otherwise: n3/5-antichain S2 ⊆ S1. <3 n1/5-chain in S2 → disjoint family Otherwise: n2/5-antichain S3 ⊆ S2. <4 n1/5-chain in S3 → disjoint family Otherwise: n1/5-antichain S4 ⊆ S3.

slide-34
SLIDE 34

n convex sets in the plane

Sets A, B are disjoint iff they are comparable in any of <1, . . . , <4.

Fact

Every N-element poset contains a t-chain or an (N/t)-antichain. <1 n1/5-chain → disjoint family Otherwise: n4/5-antichain S1. <2 n1/5-chain in S1 → disjoint family Otherwise: n3/5-antichain S2 ⊆ S1. <3 n1/5-chain in S2 → disjoint family Otherwise: n2/5-antichain S3 ⊆ S2. <4 n1/5-chain in S3 → disjoint family Otherwise: n1/5-antichain S4 ⊆ S3. ⇒ S4 incomparable in all 4 posets → intersecting family.

slide-35
SLIDE 35

Graph language

slide-36
SLIDE 36

Graph language

Comparability graph of a poset

Connect a, b with an edge if a < b or b < a.

slide-37
SLIDE 37

Graph language

Comparability graph of a poset

Connect a, b with an edge if a < b or b < a.

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

slide-38
SLIDE 38

Graph language

Comparability graph of a poset

Connect a, b with an edge if a < b or b < a.

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

convex sets n1/5

slide-39
SLIDE 39

Graph language

Comparability graph of a poset

Connect a, b with an edge if a < b or b < a.

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

convex sets n1/5 halflines n1/3

slide-40
SLIDE 40

Graph language

Comparability graph of a poset

Connect a, b with an edge if a < b or b < a.

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

convex sets n1/5 halflines n1/3

slide-41
SLIDE 41

Graph language

Comparability graph of a poset

Connect a, b with an edge if a < b or b < a.

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

convex sets n1/5 halflines n1/3

slide-42
SLIDE 42

Graph language

Comparability graph of a poset

Connect a, b with an edge if a < b or b < a.

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

convex sets n1/5 halflines n1/3

slide-43
SLIDE 43

Graph language

Comparability graph of a poset

Connect a, b with an edge if a < b or b < a.

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

convex sets n1/5 halflines n1/3

slide-44
SLIDE 44

Graph language

Comparability graph of a poset

Connect a, b with an edge if a < b or b < a.

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

Lower bound convex sets n1/5 halflines n1/3

slide-45
SLIDE 45

Graph language

Comparability graph of a poset

Connect a, b with an edge if a < b or b < a.

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

Lower bound Upper bound convex sets n1/5 n0.405 (Kynˇ cl) halflines n1/3 n0.431 (LMPT)

slide-46
SLIDE 46

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

slide-47
SLIDE 47

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

Tight for k = 1:

slide-48
SLIDE 48

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

Tight for k = 1: ∨ ∨ ∨ √n √n

slide-49
SLIDE 49

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

slide-50
SLIDE 50

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

◮ let fk(n) = max size guaranteed in any such G

slide-51
SLIDE 51

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

◮ let fk(n) = max size guaranteed in any such G

Theorem (Dumitrescu-T´

  • th, 2002)

n

1 k+1 ≤ fk(n) ≤ n 1+log k k

slide-52
SLIDE 52

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

◮ let fk(n) = max size guaranteed in any such G

Theorem (Dumitrescu-T´

  • th, 2002)

n

1 k+1 ≤ fk(n) ≤ n 1+log k k

For k = 2:

slide-53
SLIDE 53

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

◮ let fk(n) = max size guaranteed in any such G

Theorem (Dumitrescu-T´

  • th, 2002)

n

1 k+1 ≤ fk(n) ≤ n 1+log k k

For k = 2: LMPT (halflines) ⇒ f2(n) ≤ n0.431

slide-54
SLIDE 54

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

◮ let fk(n) = max size guaranteed in any such G

Theorem (Dumitrescu-T´

  • th, 2002)

n

1 k+1 ≤ fk(n) ≤ n 1+log k k

For k = 2: LMPT (halflines) ⇒ f2(n) ≤ n0.431 Dumitrescu-T´

  • th:

f2(n) ≤ n0.4118

slide-55
SLIDE 55

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

◮ let fk(n) = max size guaranteed in any such G

Theorem (Dumitrescu-T´

  • th, 2002)

n

1 k+1 ≤ fk(n) ≤ n 1+log k k

For k = 2: LMPT (halflines) ⇒ f2(n) ≤ n0.431 Dumitrescu-T´

  • th:

f2(n) ≤ n0.4118 Szab´

  • :

f2(n) ≤ n0.3878

slide-56
SLIDE 56

Lemma

If G is the union of k comparability graphs, then G contains a clique or independent set of size n

1 k+1 .

◮ let fk(n) = max size guaranteed in any such G

Theorem (Dumitrescu-T´

  • th, 2002)

n

1 k+1 ≤ fk(n) ≤ n 1+log k k

For k = 2: LMPT (halflines) ⇒ f2(n) ≤ n0.431 Dumitrescu-T´

  • th:

f2(n) ≤ n0.4118 Szab´

  • :

f2(n) ≤ n0.3878

Theorem (K-Tomon, 2019+)

f2(n) = n1/3+o(1)

slide-57
SLIDE 57

Construction for f2(n) ≤ n1/3+o(1)

slide-58
SLIDE 58

Construction for f2(n) ≤ n1/3+o(1)

n1/3

slide-59
SLIDE 59

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3

slide-60
SLIDE 60

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3

slide-61
SLIDE 61

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3

slide-62
SLIDE 62

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3

slide-63
SLIDE 63

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3

slide-64
SLIDE 64

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3

slide-65
SLIDE 65

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3

slide-66
SLIDE 66

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3

slide-67
SLIDE 67

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3

slide-68
SLIDE 68

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3 <1 <2

slide-69
SLIDE 69

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3 <1 <2 Claim 1: α(G) = n1/3

slide-70
SLIDE 70

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3 <1 <2 Claim 1: α(G) = n1/3 Claim 2: ω(G) ≤ n1/3 ·

log n log log n

slide-71
SLIDE 71

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3 <1 <2 Claim 1: α(G) = n1/3 Claim 2: ω(G) ≤ n1/3 ·

log n log log n

Claim 1:

slide-72
SLIDE 72

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3 <1 <2 Claim 1: α(G) = n1/3 Claim 2: ω(G) ≤ n1/3 ·

log n log log n

Claim 1:

  • every independent set is within a row or column
slide-73
SLIDE 73

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3 <1 <2 Claim 1: α(G) = n1/3 Claim 2: ω(G) ≤ n1/3 ·

log n log log n

Claim 1:

  • every independent set is within a row or column
  • hence it is covered by n1/3 chains
slide-74
SLIDE 74

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3 <1 <2 Claim 1: α(G) = n1/3 Claim 2: ω(G) ≤ n1/3 ·

log n log log n

Claim 2:

slide-75
SLIDE 75

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3 <1 <2 Claim 1: α(G) = n1/3 Claim 2: ω(G) ≤ n1/3 ·

log n log log n

Claim 2:

slide-76
SLIDE 76

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3 <1 <2 Claim 1: α(G) = n1/3 Claim 2: ω(G) ≤ n1/3 ·

log n log log n

Claim 2:

slide-77
SLIDE 77

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3 <1 <2 Claim 1: α(G) = n1/3 Claim 2: ω(G) ≤ n1/3 ·

log n log log n

Claim 2:

  • every clique corresponds to a fixed chain in each row and column
slide-78
SLIDE 78

Construction for f2(n) ≤ n1/3+o(1)

n1/3 n1/3 <1 <2 Claim 1: α(G) = n1/3 Claim 2: ω(G) ≤ n1/3 ·

log n log log n

Claim 2:

  • every clique corresponds to a fixed chain in each row and column
  • P(fixed row chain intersects fixed column chain) = n−1/3
slide-79
SLIDE 79

Remarks

◮ Improving n1/3 for halflines needs different approach

slide-80
SLIDE 80

Remarks

◮ Improving n1/3 for halflines needs different approach But:

slide-81
SLIDE 81

Remarks

◮ Improving n1/3 for halflines needs different approach But:

  • We could not generalize to more partial orders.
slide-82
SLIDE 82

Remarks

◮ Improving n1/3 for halflines needs different approach But:

  • We could not generalize to more partial orders.
  • Perhaps bound for convex sets could be improved?
slide-83
SLIDE 83

Remarks

◮ Improving n1/3 for halflines needs different approach But:

  • We could not generalize to more partial orders.
  • Perhaps bound for convex sets could be improved?
slide-84
SLIDE 84

Remarks

◮ Improving n1/3 for halflines needs different approach But:

  • We could not generalize to more partial orders.
  • Perhaps bound for convex sets could be improved?

Let gk(n) be max size of complete or empty bipartite graph in the union of k comparability graphs.

Theorem (Fox-Pach, 2009)

n · 2−(log log n)k ≤ gk(n) ≤ n · (log log n)k−1

(log n)k

slide-85
SLIDE 85

Remarks

◮ Improving n1/3 for halflines needs different approach But:

  • We could not generalize to more partial orders.
  • Perhaps bound for convex sets could be improved?

Let gk(n) be max size of complete or empty bipartite graph in the union of k comparability graphs.

Theorem (Fox-Pach, 2009)

n · 2−(log log n)k ≤ gk(n) ≤ n · (log log n)k−1

(log n)k

Theorem (K-Tomon, 2019+)

gk(n) ≤

n (log n)k

slide-86
SLIDE 86

Thank you!