Graphon space is compact #
The canonical graphon space GraphonSpaceI is a compact metric space (the
Lovász--Szegedy compactness theorem). It combines completeness of unit-interval graphons
with total boundedness of their cut-distance quotient. The graphon space over an atomless standard
Borel carrier is isometric to GraphonSpaceI (isometryEquivGraphonSpaceI), so it is compact too,
and hence complete.
Compactness passes to the mixing measures: ProbabilityMeasure GraphonSpaceI is compact and
metrizable (the compact-space direction of Prokhorov's theorem), so every sequence of probability
measures on the graphon space has a weakly convergent subsequence
(exists_subseq_tendsto_probabilityMeasure).
Main results #
TauCeti.DenseGraphLimits.GraphonSpaceI.instCompactSpace--GraphonSpaceIis compact.TauCeti.DenseGraphLimits.GraphonSpace.instCompactSpace-- the graphon space over an atomless standard Borel carrier is compact.TauCeti.DenseGraphLimits.exists_subseq_tendsto_probabilityMeasure-- every sequence of probability measures onGraphonSpaceIhas a weakly convergent subsequence.
References #
- L. Lovász and B. Szegedy, Szemerédi's Lemma for the Analyst, GAFA 17 (2007), Theorem 5.1.
- L. Lovász, Large Networks and Graph Limits, AMS Colloquium Publications 60 (2012), Theorem 9.23.
- S. Janson, Graphons, cut norm and distance, couplings and rearrangements, NYJM Monographs 4 (2013), Theorem A.7.
Lovász--Szegedy compactness. The cut-distance quotient of unit-interval graphons is a compact metric space. This supplies compactness for graphon-space arguments.
Lovász--Szegedy compactness over atomless standard Borel carriers. The cut-distance
quotient of graphons on a standard Borel space with an atomless probability measure is a compact
metric space, being isometric to GraphonSpaceI.
Compactness extraction. Every sequence of probability measures on GraphonSpaceI has a
weakly convergent subsequence: the graphon space is a compact metric space, so its space of
probability measures is compact and metrizable (the compact-space direction of Prokhorov's
theorem, with no tightness argument).