Which graphs are extremal Lszl Lovsz Etvs Lornd

  • Slides: 29
Download presentation
Which graphs are extremal? László Lovász Eötvös Loránd University, Budapest Joint work with Balázs

Which graphs are extremal? László Lovász Eötvös Loránd University, Budapest Joint work with Balázs Szegedy

Types of extremal graphs Given: # of nodes n, # edges m, minimize #

Types of extremal graphs Given: # of nodes n, # edges m, minimize # triangles. Triangles

Types of extremal graphs Triangles Kruskal-Katona 1 Razborov 2006 Fisher Goodman 0 1/2 2/3

Types of extremal graphs Triangles Kruskal-Katona 1 Razborov 2006 Fisher Goodman 0 1/2 2/3 3/4 1 Bollobás Mantel-Turán Lovász-Simonovits

Types of extremal graphs Triangles Given: # of nodes n, # edges m, minimize

Types of extremal graphs Triangles Given: # of nodes n, # edges m, minimize # triangles. Mantel (1907): Goodman (1959): extremal graph is Razborov (2006): extremal graph is

Types of extremal graphs Nonbipartite excluded subgraphs Given: # of nodes n, excluded subgraphs

Types of extremal graphs Nonbipartite excluded subgraphs Given: # of nodes n, excluded subgraphs F 1, . . . , Fm, minimize # edges. Erdős-Stone-Simonovits: Asymptotic extreme graph: , where

Types of extremal graphs Quadrilaterals Given: # of nodes n, # edges m, minimize

Types of extremal graphs Quadrilaterals Given: # of nodes n, # edges m, minimize # quadrilaterals. Asymptotically: Asymptotic extreme graph: random

Limits of dense graph sequences Borgs, Chayes, L, Sós, B. Szegedy, Vesztergombi Probability that

Limits of dense graph sequences Borgs, Chayes, L, Sós, B. Szegedy, Vesztergombi Probability that random map V(G) V(H) is a hom convergent: for every simple graph is convergent

The limit object as a function graphons

The limit object as a function graphons

The limit object as a function Example 1: Adjacency matrix of graph G: Examples

The limit object as a function Example 1: Adjacency matrix of graph G: Examples Associated function WG:

The limit object as a function Examples Stepfunction: Stepfunctions finite graphs with node and

The limit object as a function Examples Stepfunction: Stepfunctions finite graphs with node and edgeweights

The limit object as a function Examples Example 2: t(F, W)= 2 -|E(F)| #

The limit object as a function Examples Example 2: t(F, W)= 2 -|E(F)| # of eulerian orientations of F

The limit object as a function

The limit object as a function

Summary of main facts about graph limits For every convergent graph sequence (Gn) there

Summary of main facts about graph limits For every convergent graph sequence (Gn) there is a graphon such that Conversely, for every graphon W there is a graph sequence (Gn) such that W is essentially unique (up to measure-preserving transformation).

Extremal graphons Extremal graphon problem: Given find subject to

Extremal graphons Extremal graphon problem: Given find subject to

Finite forcing Graphon W is finitely forcible: Every finitely forcible graphon is extremal: minimize

Finite forcing Graphon W is finitely forcible: Every finitely forcible graphon is extremal: minimize Every unique extremal graphon is finitely forcible. ? ? ? Every extremal graph problem has a finitely forcible extremal graphon ? ? ?

Finite forcing Examples Goodman 1/2 Graham. Chung. Wilson

Finite forcing Examples Goodman 1/2 Graham. Chung. Wilson

Finite forcing Finitely expressible properties d-regular graphon: d-regular

Finite forcing Finitely expressible properties d-regular graphon: d-regular

Finite forcing Finitely expressible properties W is 0 -1 valued, and can be rearranged

Finite forcing Finitely expressible properties W is 0 -1 valued, and can be rearranged to be monotone decreasing in both variables “W is 0 -1 valued” is not finitely expressible. W is 0 -1 valued

Finite forcing Stepfunctions Every stepfunction is finitely forcible Is the converse true? L –

Finite forcing Stepfunctions Every stepfunction is finitely forcible Is the converse true? L – T. Sós

Finite forcing One-variable analogue (F connected) The t(F, W ) determine W up to

Finite forcing One-variable analogue (F connected) The t(F, W ) determine W up to measure preserving transformation The M(k, f ) determine f up to measure preserving transformation t(F 1, W ), . . . t(Fm, W ) are independent. M(k 1, f ), . . . M(km, f ) are independent. Characterizations: inclusion-exclusion inequalities, semidefiniteness, . . .

Finite forcing One-variable analogue W stepfunction finite number of f stepfunction finite number of

Finite forcing One-variable analogue W stepfunction finite number of f stepfunction finite number of subgraph densities that determine it. moments that determine it. ? f not a stepfunction k 1, . . . , km stepfunction g such that

Finite forcing Signs of polynomials p monotone decreasing symmetric polynomial finitely forcible ?

Finite forcing Signs of polynomials p monotone decreasing symmetric polynomial finitely forcible ?

Finite forcing Signs of polynomials, proof sketch S p(x, y)=0

Finite forcing Signs of polynomials, proof sketch S p(x, y)=0

Finite forcing 2 -dimensional graphons Is the following graphon finitely forcible? angle <π/2

Finite forcing 2 -dimensional graphons Is the following graphon finitely forcible? angle <π/2

Finite forcing UC graphons Union-complement graph (UC-graphs, cographs): Constructed from single nodes by disjoint

Finite forcing UC graphons Union-complement graph (UC-graphs, cographs): Constructed from single nodes by disjoint union and complementation G is UC no induced P 4 union and complement single node

Finite forcing UC graphons UC-graphon: not connected probability measure on paths UC-graphon is a

Finite forcing UC graphons UC-graphon: not connected probability measure on paths UC-graphon is a stepfunction tree is finite . . . connected

Finite forcing Regular UC graphons UC-graphon is regular tree is locally finite For every

Finite forcing Regular UC graphons UC-graphon is regular tree is locally finite For every locally finite tree with outdegrees ≥ 2 there is a unique regular UC-graphon. d-regular UC-graphon is a stepfunction d is rational 0 ≤ d ≤ 1 d-regular UC-graphon

Finite forcing Regular UC graphons W is not a stepfunction

Finite forcing Regular UC graphons W is not a stepfunction

Finite forcing Non-forcible graphons W(x, y) is a polynomial not finitely forcible ? ?

Finite forcing Non-forcible graphons W(x, y) is a polynomial not finitely forcible ? ? Every finitely forcible function has a finite range. ? ? (stepfunction in a weaker sense)