G-Invariance for Meta-Variational Principle #
This file defines the symmetry group G and invariance properties for the meta-variational principle framework.
Main definitions #
DirichletAutomorphism
: Automorphisms preserving Dirichlet formScaleTransform
: Scale transformations on multi-scale parametersTimeReparam
: Time reparametrizationGAction
: The full symmetry group G- Invariance predicates for various structures
Implementation notes #
The group G = Aut_ℰ(X) ⋉ (Doob × Scale × Reparam) acts on all components of the meta-variational principle, preserving the main inequality FG★.
Spectral penalty term for FG★
Equations
- Frourio.spectral_penalty_term cfg C_energy κ = κ * C_energy * Frourio.spectralSymbolSupNorm cfg ^ 2
Instances For
FG★ constant structure (simplified for G-invariance)
- C_energy : ℝ
Energy constant from FG★ inequality
Positivity constraint
Non-negativity (weaker than positivity, for compatibility)
Instances For
Meta EVI flags structure (simplified for G-invariance)
- fgstar_const : FGStarConstant
FG★ constant
- lam_base : ℝ
Base curvature parameter
- doob : DoobDegradation
Doob transform component
- lam_eff : ℝ
Effective rate
- lam_eff_eq : self.lam_eff = self.lam_base - 2 * self.doob.ε - spectral_penalty_term cfg self.fgstar_const.C_energy κ
Effective rate equation
Instances For
Technical lemma: withDensity by constant 1 is identity. This is a standard fact in measure theory.
Automorphism of X preserving the Dirichlet form ℰ
- toFun : X → X
The underlying measurable bijection
- invFun : X → X
Inverse function
- left_inv : Function.LeftInverse self.invFun self.toFun
Left inverse property
- right_inv : Function.RightInverse self.invFun self.toFun
Right inverse property
- measurable_toFun : Measurable self.toFun
Measurability of forward map
- measurable_invFun : Measurable self.invFun
Measurability of inverse map
- preserves_dirichlet (Γ : CarreDuChamp X) (f g : X → ℝ) : (Γ.Γ (fun (x : X) => f (self.toFun x)) fun (x : X) => g (self.toFun x)) = fun (x : X) => Γ.Γ f g (self.toFun x)
Dirichlet-form invariance for every CarreDuChamp Γ: Pullback by
toFun
commutes with Γ.
Instances For
Apply scale transformation to configuration
Equations
Instances For
Time reparametrization for curves
The reparametrization function θ : [0,1] → [0,1]
Monotone increasing
Boundary conditions
- continuous : Continuous self.θ
Continuity of the reparametrization function
Instances For
The symmetry group G for the meta-variational principle. G = Aut_ℰ(X) ⋉ (Doob × ℝ₊ × Reparam)
- aut : DirichletAutomorphism
Dirichlet form automorphism component
- doob_h : X → ℝ
Doob transform component (h function)
Positivity of Doob function
- scale : ScaleTransform m
Scale transformation component
- reparam : TimeReparam
Time reparametrization component
Instances For
Action of G on measures
Equations
- g.actOnMeasure μ = (MeasureTheory.Measure.map g.aut.toFun μ).withDensity fun (x : X) => ENNReal.ofReal (g.doob_h x ^ 2)
Instances For
Action of G on multi-scale configuration
Equations
- g.actOnConfig cfg = g.scale.apply cfg
Instances For
Invariance of entropy under G-action (up to additive constant)
Equations
- Frourio.entropy_G_invariant Ent = ∀ (g : Frourio.GAction X m) (ρ : MeasureTheory.Measure X), ∃ (c : ℝ), Ent (g.actOnMeasure ρ) = Ent ρ + c
Instances For
Invariance of modified distance under G-action
Equations
- One or more equations did not get rendered due to their size.
Instances For
Invariance of multi-scale difference operator under pullback
Equations
- One or more equations did not get rendered due to their size.
Instances For
Scale invariance of spectral symbol sup-norm. Under scale transformation τ ↦ στ with spectral variable λ ↦ λ/σ, the sup-norm ‖ψ_m‖_∞ is exactly preserved.
Pushforward of a measure by a measurable function
Equations
- Frourio.pushforward f _hf μ = MeasureTheory.Measure.map f μ
Instances For
Entropy is invariant under pushforward by measure-preserving maps
Distance compatibility under pullback/pushforward
Carré du Champ operator compatibility with pullback
Entropy functional with pushforward/pullback structure
- μ : MeasureTheory.Measure X
Base measure
- Ent : MeasureTheory.Measure X → ENNReal
Entropy relative to base measure
- pushforward_compat (f : X → X) (_hf : Measurable f) : self.Ent (MeasureTheory.Measure.map f self.μ) = 0
Pushforward compatibility
- pullback_integrable (φ : X → ℝ) (g : DirichletAutomorphism) : MeasureTheory.Integrable φ self.μ → MeasureTheory.Integrable (pullback g.toFun φ) self.μ
Pullback of test functions preserves integrability
Instances For
Modified distance with pullback/pushforward structure
- μ : MeasureTheory.Measure X
Base measure
- d_m : MeasureTheory.Measure X → MeasureTheory.Measure X → ℝ
The modified distance function
- pushforward_preserves (g : DirichletAutomorphism) (ρ₀ ρ₁ : MeasureTheory.Measure X) : self.d_m (pushforward g.toFun ⋯ ρ₀) (pushforward g.toFun ⋯ ρ₁) = self.d_m ρ₀ ρ₁
Pushforward preserves distance
- pullback_velocity : DirichletAutomorphism → (X → ℝ) → X → ℝ
Pullback of velocity potentials
Instances For
Theorem: Entropy transformation under pullback/pushforward
Distance formula under pullback/pushforward
Pullback preserves L² integrability
Main theorem: Complete G-invariance of the meta-variational principle
Main G-invariance for FG★ (effective rate).
If the Doob degradation amount ε
is fixed (encoded in doob : DoobDegradation
),
then the effective rate λ_eff = (λ - 2ε) - κ C ‖ψ_m‖_∞^2
is invariant under the
G
-action components (Dirichlet automorphism, Doob, scale, reparam).
Composition of G-actions forms a group structure
Equations
- One or more equations did not get rendered due to their size.
Instances For
Identity element of G-action
Equations
- One or more equations did not get rendered due to their size.
Instances For
Identity action on measures: doing nothing leaves the measure unchanged.
Identity action on configurations: parameters remain unchanged.
Invariance of entropy under Dirichlet automorphism. The key observation is that when Doob h ≡ 1, the measure transformation reduces to a simple pushforward.
Scale transformation law for dm. Under scale transformation τ ↦ στ, the modified distance transforms as: dm(στ) = σ · dm(τ) when κ is scaled appropriately.
Invariance of dm under full G-action with scale component. When all components (measures and parameters) are transformed consistently, the distance is preserved.
Main theorem: Full G-invariance of meta-variational principle
Corollary: The effective rate λ_eff is G-invariant up to Doob composition