random graphs, finite - pennsylvania state universitya random‑free graphon is countably universal...

23
Random graphs, finite extension constructions, and complexity Jan Reimann, Penn State October 4, 2015 joint work with Cameron Freer, MIT

Upload: others

Post on 19-May-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

Random graphs, finiteextension constructions, and

complexityJan Reimann, Penn State 

October 4, 2015

joint work with Cameron Freer, MIT

Page 2: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

Homogeneous structuresA countable (relational) structure   is homogeneous ifevery isomorphism between finite substructures of extends to an automorphism of  .

Fraissé: Any homogeneous structure arises as aamalgamation process of finite structures over the samelanguage (Fraissé limits).

Examples:

,the Rado (random) graphthe universal  ‑free graphs,   (Henson)

(ℚ, <)

Kn n ~ 3

Page 3: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

Randomized constructionsMany homogeneous structures can obtained (almostsurely) by adding new points according to a randomizedprocess.

: add the  ‑th point between (or at the ends) ofany existing point with uniform probability  .

Rado graph: add the  ‑th vertex and connect to everyprevious vertex with probability   (uniformly andindependently).

Vershik (2004): Urysohn space, Droste and Kuske (2003): universal poset

Henson graph: ??? (until 2008)

(ℚ, <) n1/n

np

Page 4: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

Constructions ʺfrom belowʺA naive approach to ʺrandomizeʺ the construction of theHenson graph would be as follows:

In the  ‑th step of the construction, pick a one‑vertexextension uniformly among all possible extensions thatpreserve  ‑freeness.

However: Erdös, Kleitman, and Rothschild (1976)showed that this asymptotically almost surely yields abipartite graph (in fact, the universal countable bipartitegraph).

The Henson graph(s), in contrast, has to contain and hence cannot be bipartite.

n

Kn

C5

Page 5: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

Symmetric constructionsOn the other hand, one could (degenerately) ensure thatevery triangle‑free subgraph appears, and indeed witnessall extension axioms, by deterministically building theHenson graph.

But this violates symmetry: we would like the jointdistribution of any distinct k‑tuple to be the same as anyother (i.e., exchangeability).

from Lloyd–Orbanz–Ghahramani–Roy (2012), Random function priors for exchangeable arrays

Page 6: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

from Lloyd–Orbanz–Ghahramani–Roy (2012), Random function priors for exchangeable arrays

Symmetric constructionsIs there an exchangeable construction of the Hensongraph?

Equivalently, is there a probability measure on graphswith vertex set   that is concentrated on the isomorphismclass of the Henson graph, and is invariant under thelogic action of the symmetric group   on the underlyingset   of vertices?

ω

S7ω

Page 7: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

Constructions ʺfrom aboveʺPetrov and Vershik (2010) showed how to constructuniversal  ‑free graphs probabilistically by samplingthem from a continuous graph.

Indeed every exchangeable structure in a countablelanguage must arise in essentially this way, as shown byAldous (1981) and Hoover (1979).

These continuous graphs, known as graphons, have beenstudied extensively over the past decade.

See, for example the recent book by Lovasz, Largenetworks and graph limits (2012).

Kn

Page 8: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

GraphonsOne basic motivation behind graphons is to capture theasymtotic behavior of growing sequences of densegraphs, e.g. with respect to subgraph densities.

While the Rado graph can be seen as the limit object of asequence   of finite random graphs, it does notdistinguish between the distributions with which theedges are produced.

For any  ,   ʺconvergesʺ almost surely to(an isomorphic copy of) the Rado graph.

However, if  ,   will exhibit subgraphdensities very different from 

( )Gn

0 < p < 1 �(n, p)

�p1 p2 �(n, )p1

�(n, )p2

Page 9: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

ConvergenceLet   be a graph sequence with  .

We say   converges if

( )Gn |V( )| → 7Gn

( )Gn

for every finite graph  , the relative number  of embeddings of   into 

converges.

F(F, )ti Gn F Gn

Page 10: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

Graphons measurable, and for all  , 

 and  .

Think:   is the probability there is an edge between  and  .

Subgraph densities:

edges: triangles: this can be generalized to define  .

W : [0, 1 → [0, 1]]2 x, y

W(x, x) = 0 W(x, y) = W(y, x)

W(x, y)x y

D W(x, y) dx dyD W(x, y)W(y, z)W(z, x) dx dy dz

(F, W)ti

Page 11: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

Graphons and graph limitsBasic idea: ʺpixel picturesʺ

from Lovasz (2012), Large networks and graph limits

Page 12: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

Graphons and graph limitsConvergence of pixel pictures

from Lovasz (2012), Large networks and graph limits

Page 13: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

Graphons and graph limitsConvergence of pixel pictures

from Lovasz (2012), Large networks and graph limits

Page 14: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

The limit graphonTHM: For every convergent graph sequence   thereexists (up to weak isomorphim) exactly one graphon 

such that for all finite  :

( )Gn

WF

.(F, ) ⟶ (F, W)ti Gn ti

Page 15: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

Sampling from graphonsWe can obtain a finite graph   from   by(independently) sampling   points   from and filling edges according to probabilities  .

almost surely, we get a sequence with  .

If we sample  ‑many points from  , wealmost surely get the random graph.

�(n, W) Wn ,… ,x1 xn [0, 1]

W( , )xi xj

�(n, W) → W

ω W(x, y) z 1/2

Page 16: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

The Petrov‑Vershik graphonPetrov and Vershik (2010) constructed, for each  , agraphon   such that we almost surely sample a Hensongraph for  .

The graphons are (necessarily) {0,1}‑valued.

Such graphons are called random‑free.

The constructions resembles a finite extensionconstruction with simple geometric forms, where eachstep satisfies a new type requiring attention.

The method can also be used to construct random‑freegraphons from which we sample the Rado graph.

n ~ 3Wn

Page 17: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

Invariant measuresThe Petrov‑Vershik graphon also yields a measure on theset of countable infinite graphs concentrating on the set ofuniversal, homogeneous  ‑free graphs.

This measure will be invariant under the ʺlogic actionʺ,the natural action of   on the space of countable(relational) structures with universe  .

This method was generalized by Ackerman, Freer, and Patel(2014) to other homogeneous structures.

It can be used to define algorithmic randomness for suchstructures (as suggested by Nies and Fouché).

Kn

S7ℕ

Page 18: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

Universal graphonsA random‑free graphon is countably universal if for everyset of distinct points from  ,  ,  ,the intersection

has non‑empty interior. Here   is the neighborhood of  .

For countably  ‑free universal graphs, we require this tohold only for such tuples where the induced subgraph bythe   has no induced  ‑subgraph,also require that no n‑tuples induce a  .

[0, 1] , ,… ,x1 x2 xn ,… ,y1 ym

B⋂i,j

Exi ECyj

= {y: W(x, y) = 1}Ex x

Kn

xi Kn+1

Kn

Page 19: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

The topology of graphonsNeighborhood distance: 

and mod out by  .

Example:   is a singleton space.

THM: (Freer & R.) (informal) If   is a random‑freeuniversal graphon obtained via a ʺtameʺ extensionmethod, then   is not compact in the   topology.

(x, y) = > W(x, . ) + W(y, . ) = ∫ |W(x, z) + W(y, z)| dzrW >1

(x, y) = 0rW

W(x, y) z p

W

W rW

Page 20: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

ʺTameʺ extensionsDEF: A random‑free graphon   has continuous realizationof extensions if there exists a function

that is continuous a.e. such that for all  ,

Here   is the neighborhood of  .  The Petrov‑Vershik graphons have uniformly continuousrealization of extensions.

W

f : ( ,… , ), ( ,… , ) ↦ (l, r)x1 xn y1 ym

,x ⃗ y ⃗

[l, r] � B .⋂i,j

Exi ECyj

= {y: W(x, y) = 1}Ex x

Page 21: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

Non‑compactness  

THM: If a countably ( ‑free) universalgraphon has uniformly continuous realizationof extensions, then it is not compact in the 

‑topology.

Kn

rW

Page 22: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

Compactness of graphonsThis contrasts the following result due to Lovasz andSzegedy.

THM: If a pure graphon   misses some signedbipartite graph  , then

(i)   is compact, and (ii) has Minkowski dimension at most  .

(J, W)F

(J, )rW

10v(F)

Page 23: Random graphs, finite - Pennsylvania State UniversityA random‑free graphon is countably universal if for every set of distinct points from , , , the intersection has non‑empty

Complexity of universal graphonsconstruction: fully

randomtamedeterministic

generaldeterministic

complexityof graphon

low 

(singleton)

high 

(not compact,infiniteMinkowskidimension)

           ?