Skip to content

Monotone couplings and stochastic domination

5 fully proved

Stat Mech

DefinitionSet of coordinatewise ordered configuration pairs#✓fully proved

The set of ordered pairs (ω, ω') of configurations with ω ≤ ω' coordinatewise.

def orderSet (E : Type*) : Set (ConfigSpace E × ConfigSpace E) := {p | p.1 ≤ p.2}
Uses 1 · Used by 6

StructureMonotone coupling of two measures#✓fully proved

κ is a monotone coupling of μ and ν: a measure on pairs of configurations with first marginal μ, second marginal ν, and no mass off the ordered pairs ω ≤ ω'.

structure MonotoneCoupling (μ ν : Measure (ConfigSpace E))
    (κ : Measure (ConfigSpace E × ConfigSpace E)) : Prop where
  /-- The first marginal of `κ` is `μ`. -/
  fst_eq : κ.fst = μ
  /-- The second marginal of `κ` is `ν`. -/
  snd_eq : κ.snd = ν
  /-- `κ` is concentrated on the ordered pairs. -/
  supported : κ (orderSet E)ᶜ = 0
Uses 2 · Used by 8

Monotone Coupling

TheoremMarginals of a coupling have equal total mass#✓fully proved

The two marginals of a coupling have the same total mass.

theorem total_mass_eq (h : MonotoneCoupling μ ν κ) : μ Set.univ = ν Set.univ
Uses 2

TheoremStrassen's theorem, easy direction#✓fully proved

A finite monotone coupling of μ and ν shows μ ≼ ν: the easy direction of Strassen's theorem.

theorem stochasticallyDominated (h : MonotoneCoupling μ ν κ)
    [IsFiniteMeasure κ] : μ ≼ ν
Uses 5 · Used by 1

Stat Mech

TheoremExistence of a monotone coupling implies stochastic domination#✓fully proved

If a finite monotone coupling of μ and ν exists, then μ ≼ ν.

theorem stochasticallyDominated_of_exists_monotoneCoupling
    {μ ν : Measure (ConfigSpace E)}
    (h : ∃ κ : Measure (ConfigSpace E × ConfigSpace E),
          IsFiniteMeasure κ ∧ MonotoneCoupling μ ν κ) :
    μ ≼ ν
Uses 4

The dependency explorer needs JavaScript and its graph data.

Download the complete graph data.