Graph uniformity and uniform partitions #
In this file we define uniformity of a pair of vertices in a graph and uniformity of a partition of vertices of a graph. Both are also known as ε-regularity.
Finsets of vertices s and t are ε-uniform in a graph G if their edge density is at most
ε-far from the density of any big enough s' and t' where s' ⊆ s, t' ⊆ t.
The definition is pretty technical, but it amounts to the edges between s and t being "random"
The literature contains several definitions which are equivalent up to scaling ε by some constant
when the partition is equitable.
A partition P of the vertices is ε-uniform if the proportion of non ε-uniform pairs of parts
is less than ε.
Main declarations #
SimpleGraph.IsUniform: Graph uniformity of a pair of finsets of vertices.SimpleGraph.nonuniformWitness:G.nonuniformWitness ε s tandG.nonuniformWitness ε t stogether witness the non-uniformity ofsandt.Finpartition.nonUniforms: Non uniform pairs of parts of a partition.Finpartition.IsUniform: Uniformity of a partition.Finpartition.nonuniformWitnesses: For each non-uniform pair of parts of a partition, pick witnesses of non-uniformity and dump them all together.
References #
[Yaël Dillies, Bhavik Mehta, Formalising Szemerédi’s Regularity Lemma in Lean][srl_itp]
Graph uniformity #
A pair of finsets of vertices is ε-uniform (aka ε-regular) iff their edge density is close
to the density of any big enough pair of subsets. Intuitively, the edges between them are
random-like.
Equations
Instances For
An arbitrary pair of subsets witnessing the non-uniformity of (s, t). If (s, t) is uniform,
returns (s, t). Witnesses for (s, t) and (t, s) don't necessarily match. See
SimpleGraph.nonuniformWitness.
Equations
Instances For
Arbitrary witness of non-uniformity. G.nonuniformWitness ε s t and
G.nonuniformWitness ε t s form a pair of subsets witnessing the non-uniformity of (s, t). If
(s, t) is uniform, returns s.
Equations
- G.nonuniformWitness ε s t = if WellOrderingRel s t then (G.nonuniformWitnesses ε s t).1 else (G.nonuniformWitnesses ε t s).2
Instances For
Uniform partitions #
The pairs of parts of a partition P which are not ε-dense in a graph G. Note that we
dismiss the diagonal. We do not care whether s is ε-dense with itself.
Equations
- P.sparsePairs G ε = {x ∈ P.parts.offDiag | match x with | (u, v) => ↑(G.edgeDensity u v) < ε}
Instances For
The pairs of parts of a partition P which are not ε-uniform in a graph G. Note that we
dismiss the diagonal. We do not care whether s is ε-uniform with itself.
Instances For
A finpartition of a graph's vertex set is ε-uniform (aka ε-regular) iff the proportion of
its pairs of parts that are not ε-uniform is at most ε.
Instances For
A choice of witnesses of non-uniformity among the parts of a finpartition.
Equations
- P.nonuniformWitnesses G ε s = Finset.image (G.nonuniformWitness ε s) ({t ∈ P.parts | s ≠ t ∧ ¬G.IsUniform ε s t})
Instances For
Equipartitions #
Reduced graph #
The reduction of the graph G along partition P has edges between ε-uniform pairs of parts
that have edge density at least δ.
Equations
- One or more equations did not get rendered due to their size.