Norm-topology LDP adapter for a sequence of laws
DefinitionTFIMPhaseDiagram_Classical_Supplement_deviationsFor probability laws on the Hilbert space and a lower semicontinuous nonnegative rate , this adapter uses the published general metric large deviation principle at speed :
for every norm-open and norm-closed . The laws are realized as coordinates of their canonical independent product, so the published random-element interface applies without changing any law. Nonnegativity, lower semicontinuity and the probability-law requirements are part of the predicate, rather than additional hypotheses of a source theorem.
Formalization Note Coordinate has source index , with speed . The index shift removes the irrelevant zero-speed first term. The published definition uses extended logarithms with and extended infima; conversion of rates to extended nonnegative reals is accompanied by the nonnegativity clause. This reuses the general LDP object rather than redeclaring its bounds.
import Definitions.Def_TFIMPhaseDiagram_Classical_Supplement_pathSpace
import Definitions.Def_ReflectedRWLDP_SamplePath_LDP
set_option autoImplicit false
open MeasureTheory Filter
open scoped ENNReal
noncomputable section
namespace TFIMPhaseDiagram.Classical.Supplement
/-- Law-sequence adapter to the published general metric LDP, using the canonical independent
coordinate realization. Index k is source index k+1, so the speed is k+1, never zero.
The rate is required to be lower semicontinuous and nonnegative, as in footnote 10. -/
def HasNormLDP (μ : ℕ → Measure CircleL2) (I : CircleL2 → EReal) : Prop :=
LowerSemicontinuous I ∧ (∀ x, 0 ≤ I x) ∧
∃ hp : ∀ k : ℕ, IsProbabilityMeasure (μ (k+1)),
letI : ∀ k : ℕ, IsProbabilityMeasure (μ (k+1)) := hp
ReflectedRWLDP.SamplePath.SatisfiesLDP
(Measure.infinitePi (fun k => μ (k+1))) (fun k σ => σ k)
edist (fun k => (k+1 : ℕ)) (fun x => (I x).toENNReal)
end TFIMPhaseDiagram.Classical.Supplement