OAI/Analysis/Brenier/Translation.lean
19 declarations, 0 contain sorry, 0 declared as axioms
- theorem
OAI.Problem358.integral_increment_eq_boundaryline 12Source documentation
Integrating a measure-preserving increment over a common core leaves only boundary terms.
theorem integral_increment_eq_boundary {X : Type*} [MeasurableSpace X] {μ : Measure X} {f : X → X} (hf : MeasurePreserving f μ μ) (he : MeasurableEmbedding f) {K A : Set X} (hA : MeasurableSet A) (hAK : A ⊆ K) (hBK : f '' A ⊆ K) {g : X → ℝ} (hg : IntegrableOn g K μ) : (∫ x in A, (g (f x) - g x) ∂μ) = (∫ x in K \ A, g x ∂μ) - ∫ x in K \ (f '' A), g x ∂μ := by … - theorem
OAI.Problem358.abs_integral_increment_le_boundaryline 28Source documentation
A bounded function has a small mean increment whenever the common core loses little mass.
theorem abs_integral_increment_le_boundary {X : Type*} [MeasurableSpace X] {μ : Measure X} {f : X → X} (hf : MeasurePreserving f μ μ) (he : MeasurableEmbedding f) {K A : Set X} (hKfin : μ K ≠ ⊤) (hA : MeasurableSet A) (hAK : A ⊆ K) (hBK : f '' A ⊆ K) {g : X → ℝ} (hg : IntegrableOn g K μ) {L : ℝ} (hbound : ∀ x ∈ K, |g x| ≤ L) : |∫ x in A, (g (f x) - g x) ∂μ| ≤ 2 * L * μ.real (K \ A) := by … - theorem
OAI.Problem358.integral_error_sq_le_boundaryline 59Source documentation
A one-sided bounded error inherits an integrated square bound from boundary mass.
theorem integral_error_sq_le_boundary {X : Type*} [MeasurableSpace X] {μ : Measure X} {f : X → X} (hf : MeasurePreserving f μ μ) (he : MeasurableEmbedding f) {K A : Set X} (hKfin : μ K ≠ ⊤) (hA : MeasurableSet A) (hAK : A ⊆ K) (hBK : f '' A ⊆ K) {g δ : X → ℝ} (hg : IntegrableOn g K μ) {L : ℝ} (hL : 0 ≤ L) (hbound : ∀ x ∈ K, |g x| ≤ L) (hδint : IntegrableOn (fun x => δ x ^ 2) A μ) (hδ : ∀ᵐ x ∂μ.restrict A, 0 ≤ δ x ∧ δ x ≤ g (f x) - g x ∧ g (f x) - g x ≤ 2 * L) : (∫ x in A, δ x ^ 2 ∂μ) ≤ 4 * L ^ 2 * μ.real (K \ A) := by … - theorem
OAI.Problem358.slice_translation_identityline 95Source documentation
Translation cancellation on a one-dimensional slice.
theorem slice_translation_identity {g : ℝ → ℝ} {a b h : ℝ} (hh : 0 ≤ h) (hhab : h ≤ b - a) (hg : IntervalIntegrable g volume a b) : (∫ t in a..b-h, (g (t+h) - g t)) = (∫ t in b-h..b, g t) - ∫ t in a..a+h, g t := by … - theorem
OAI.Problem358.slice_translation_leline 112Source documentation
Bounded derivatives have shift discrepancy at most twice the boundary strip size.
theorem slice_translation_le {g : ℝ → ℝ} {a b h L : ℝ} (hh : 0 ≤ h) (hhab : h ≤ b-a) (hg : IntervalIntegrable g volume a b) (hbound : ∀ t ∈ Icc a b, |g t| ≤ L) : (∫ t in a..b-h, (g (t+h) - g t)) ≤ 2*L*h := by … - theorem
OAI.Problem358.slice_error_sq_leline 135Source documentation
The one-dimensional squared finite-difference error estimate used on convex slices.
theorem slice_error_sq_le {g δ : ℝ → ℝ} {a b h L : ℝ} (hh : 0 ≤ h) (hhab : h ≤ b-a) (hL : 0 ≤ L) (hg : IntervalIntegrable g volume a b) (hbound : ∀ t ∈ Icc a b, |g t| ≤ L) (hδ : ∀ t ∈ Icc a (b-h), 0 ≤ δ t ∧ δ t ≤ g (t+h)-g t) (hδint : IntervalIntegrable (fun t => δ t ^ 2) volume a (b-h)) : (∫ t in a..b-h, δ t ^ 2) ≤ 4*L^2*h := by … - theorem
OAI.Problem358.slice_error_sq_integrableline 168Source documentation
Bounded measurable slice errors have an integrable square.
theorem slice_error_sq_integrable {δ : ℝ → ℝ} {a b L : ℝ} (hab : a ≤ b) (hL : 0 ≤ L) (hδmeas : Measurable δ) (hδ : ∀ t ∈ Icc a b, 0 ≤ δ t ∧ δ t ≤ 2*L) : IntervalIntegrable (fun t => δ t ^ 2) volume a b := by … - theorem
OAI.Problem358.slice_error_sq_le_of_measurableline 184Source documentation
Measurable form of the squared finite-difference error estimate.
theorem slice_error_sq_le_of_measurable {g δ : ℝ → ℝ} {a b h L : ℝ} (hh : 0 ≤ h) (hhab : h ≤ b-a) (hL : 0 ≤ L) (hg : IntervalIntegrable g volume a b) (hbound : ∀ t ∈ Icc a b, |g t| ≤ L) (hδmeas : Measurable δ) (hδ : ∀ t ∈ Icc a (b-h), 0 ≤ δ t ∧ δ t ≤ g (t+h)-g t) : (∫ t in a..b-h, δ t ^ 2) ≤ 4*L^2*h := by … - theorem
OAI.Problem358.slice_error_sq_integrable_aeline 202Source documentation
Almost-everywhere bounds suffice to establish square integrability on a slice.
theorem slice_error_sq_integrable_ae {δ : ℝ → ℝ} {a b L : ℝ} (hab : a ≤ b) (hL : 0 ≤ L) (hδmeas : AEStronglyMeasurable δ (volume.restrict (Icc a b))) (hδ : ∀ᵐ t ∂volume.restrict (Icc a b), 0 ≤ δ t ∧ δ t ≤ 2*L) : IntervalIntegrable (fun t => δ t ^ 2) volume a b := by … - theorem
OAI.Problem358.slice_error_sq_le_aeline 214Source documentation
The quotient error estimate tolerates the null exceptional set of derivative existence.
theorem slice_error_sq_le_ae {g δ : ℝ → ℝ} {a b h L : ℝ} (hh : 0 ≤ h) (hhab : h ≤ b-a) (hL : 0 ≤ L) (hg : IntervalIntegrable g volume a b) (hbound : ∀ t ∈ Icc a b, |g t| ≤ L) (hδmeas : AEStronglyMeasurable δ (volume.restrict (Icc a (b-h)))) (hδ : ∀ᵐ t ∂volume.restrict (Icc a (b-h)), 0 ≤ δ t ∧ δ t ≤ g (t+h)-g t) : (∫ t in a..b-h, δ t ^ 2) ≤ 4*L^2*h := by … - theorem
OAI.Problem358.intervalIntegrable_of_measurable_abs_boundline 262 - theorem
OAI.Problem358.intervalIntegrable_deriv_of_lipschitzline 273Source documentation
A globally Lipschitz real function has an interval-integrable derivative, including at the points where
derivis defined to be zero.theorem intervalIntegrable_deriv_of_lipschitz {u : ℝ → ℝ} {L : ℝ≥0} (hu : LipschitzWith L u) (a b : ℝ) : IntervalIntegrable (deriv u) volume a b := by … - theorem
OAI.Problem358.abs_forwardQuotient_le_of_lipschitzline 281Source documentation
Forward quotients of a Lipschitz function have the same bound.
theorem abs_forwardQuotient_le_of_lipschitz {u : ℝ → ℝ} {L : ℝ≥0} (hu : LipschitzWith L u) {h : ℝ} (hh : 0 < h) (x : ℝ) : |forwardQuotient u h x| ≤ L := by …Mentions: OAI.Problem358.forwardQuotient
- theorem
OAI.Problem358.intervalIntegrable_forwardQuotient_error_sqline 291Source documentation
Squared finite-difference errors are integrable without any continuity assumption on the derivative.
theorem intervalIntegrable_forwardQuotient_error_sq {u : ℝ → ℝ} {L : ℝ≥0} (hu : LipschitzWith L u) {h : ℝ} (hh : 0 < h) (a b : ℝ) : IntervalIntegrable (fun x => (forwardQuotient u h x - deriv u x) ^ 2) volume a b := by …Mentions: OAI.Problem358.forwardQuotient
- theorem
OAI.Problem358.ae_convex_forwardQuotient_error_sq_leline 314Source documentation
Convex quotient control holds almost everywhere for globally convex Lipschitz functions, because translation preserves null sets.
theorem ae_convex_forwardQuotient_error_sq_le {u : ℝ → ℝ} {L : ℝ≥0} (hu : ConvexOn ℝ univ u) (hLip : LipschitzWith L u) {h : ℝ} (hh : 0 < h) : ∀ᵐ x : ℝ, (forwardQuotient u h x - deriv u x) ^ 2 ≤ 2 * (L : ℝ) * (deriv u (x + h) - deriv u x) := by …Mentions: OAI.Problem358.forwardQuotient
- theorem
OAI.Problem358.convex_slice_quotient_error_sq_leline 329Source documentation
The squared quotient error integrated on the allowed part of a slice is bounded by
4 L² h, with no everywhere-differentiability assumption.theorem convex_slice_quotient_error_sq_le {u : ℝ → ℝ} {L : ℝ≥0} (hu : ConvexOn ℝ univ u) (hLip : LipschitzWith L u) {a b h : ℝ} (hh : 0 < h) (hhab : h ≤ b - a) : (∫ x in a..b-h, (forwardQuotient u h x - deriv u x) ^ 2) ≤ 4 * (L : ℝ) ^ 2 * h := by …Mentions: OAI.Problem358.forwardQuotient
- theorem
OAI.Problem358.sq_deriv_sub_le_of_lipschitzline 357Source documentation
The difference of derivatives of two Lipschitz real functions is bounded pointwise, including the default values at nondifferentiability points.
theorem sq_deriv_sub_le_of_lipschitz {u v : ℝ → ℝ} {L₀ L₁ : ℝ≥0} (hu : LipschitzWith L₀ u) (hv : LipschitzWith L₁ v) (x : ℝ) : (deriv u x - deriv v x) ^ 2 ≤ ((L₀ : ℝ) + (L₁ : ℝ)) ^ 2 := by … - theorem
OAI.Problem358.intervalIntegrable_sq_deriv_sub_of_lipschitzline 370Source documentation
Squared derivative differences are interval-integrable.
theorem intervalIntegrable_sq_deriv_sub_of_lipschitz {u v : ℝ → ℝ} {L₀ L₁ : ℝ≥0} (hu : LipschitzWith L₀ u) (hv : LipschitzWith L₁ v) (a b : ℝ) : IntervalIntegrable (fun x => (deriv u x - deriv v x) ^ 2) volume a b := by … - theorem
OAI.Problem358.integral_sq_deriv_sub_le_of_lipschitzline 382Source documentation
Boundary-strip estimate for squared derivative differences.
theorem integral_sq_deriv_sub_le_of_lipschitz {u v : ℝ → ℝ} {L₀ L₁ : ℝ≥0} (hu : LipschitzWith L₀ u) (hv : LipschitzWith L₁ v) {a b : ℝ} (hab : a ≤ b) : (∫ x in a..b, (deriv u x - deriv v x) ^ 2) ≤ ((L₀ : ℝ) + (L₁ : ℝ)) ^ 2 * (b - a) := by …