Prob split uniform sampling of equiv prod
prob_split_uniform_sampling_of_equiv_prod
Project documentation
Generic Probability Splitting Lemma (via Equivalence) This lemma proves that a single probability statement over a uniform distribution on a type α can be rewritten as a sequential probability statement over two smaller, independent distributions γ and δ, given an equivalence e : α ≃ γ × δ. This is a formal "change of variables" for probabilit...
Exact Lean statement
theorem prob_split_uniform_sampling_of_equiv_prod {α γ δ : Type}
-- Fintype & Nonempty assumptions for all types
[Fintype α] [Fintype γ] [Fintype δ]
[Nonempty α] [Nonempty γ] [Nonempty δ]
-- The equivalence that splits α into γ × δ
(e : α ≃ γ × δ)
-- The predicate on the original (combined) type
(P : α → Prop)
:
-- LHS: Probability over the combined space
Pr_{ let r ← $ᵖ α }[ P r ] =
-- RHS: Probability over the sequential, split spaces
Pr_{ let x ← $ᵖ γ; let y ← $ᵖ δ }[ P (e.symm (x, y)) ]Formal artifact
Lean source
theorem prob_split_uniform_sampling_of_equiv_prod {α γ δ : Type} -- Fintype & Nonempty assumptions for all types [Fintype α] [Fintype γ] [Fintype δ] [Nonempty α] [Nonempty γ] [Nonempty δ] -- The equivalence that splits α into γ × δ (e : α ≃ γ × δ) -- The predicate on the original (combined) type (P : α → Prop) : -- LHS: Probability over the combined space Pr_{ let r ← $ᵖ α }[ P r ] = -- RHS: Probability over the sequential, split spaces Pr_{ let x ← $ᵖ γ; let y ← $ᵖ δ }[ P (e.symm (x, y)) ] := by classical -- 1. Unroll the LHS (a single `let`) using `prStx_unfold_final` -- LHS = ∑' r, Pr[r] * (if P r then 1 else 0) rw [prob_tsum_form_singleton] let D_rest := fun (x : γ) => (do let y ← $ᵖ δ return (P (e.symm (x, y))) ) conv_rhs => apply prob_tsum_form_split_first (D := $ᵖ γ) (D_rest := D_rest) simp_rw [D_rest] simp only [PMF.uniformOfFintype_apply, mul_ite, mul_one, mul_zero] simp_rw [prob_tsum_form_singleton] -- ⊢ (∑' (x : α), ... = ∑' (x : γ), (↑(Fintype.card γ))⁻¹ * ∑' (x_1 : δ), ... conv_rhs => enter [1, x]; rw [←ENNReal.tsum_mul_left] -- ⊢ (∑' (x : α), ... = ∑' (x : γ) (i : δ), ... rw [←ENNReal.tsum_prod] -- ⊢ (∑' (x : α), ...) = (∑' (p : γ × δ), ...) conv_lhs => rw [tsum_eq_sum (α := ENNReal) (β := α) (f := fun x => if P x then (↑(Fintype.card α))⁻¹ else 0) (s := Finset.univ) (hf := fun b => by simp only [mem_univ, not_true_eq_false, ite_eq_right_iff, ENNReal.inv_eq_zero, IsEmpty.forall_iff] )] conv_rhs => rw [tsum_eq_sum (α := ENNReal) (β := γ × δ) (f := fun x => (↑(Fintype.card γ))⁻¹ * (($ᵖ δ) x.2 * if P (e.symm x) then 1 else 0) ) (s := Finset.univ) (hf := fun b => by simp only [mem_univ, not_true_eq_false, IsEmpty.forall_iff] )] -- ⊢ (∑ b : α, .. = ..) = (∑ b : γ × δ, ..) have hcard_of_equiv: (Fintype.card α) = (Fintype.card (γ × δ)) := Fintype.card_congr e rw [Finset.sum_equiv (s := Finset.univ (α := α)) (t := Finset.univ (α := γ × δ)) (f := fun x => if P x then (↑(Fintype.card α))⁻¹ else 0) (g := fun x => (↑(Fintype.card γ))⁻¹ * (($ᵖ δ) x.2 * if P (e.symm x) then 1 else 0)) (e := e) (hst := fun i => by simp only [mem_univ] ) (hfg := fun i => by simp only [mem_univ, PMF.uniformOfFintype_apply, Equiv.symm_apply_apply, mul_ite, mul_one, mul_zero, forall_const] by_cases hP : P i · simp only [hP, ↓reduceIte] rw [hcard_of_equiv] rw [ENNReal.mul_inv_rev_ENNReal (ha := Fintype.card_ne_zero)] rw [Fintype.card_prod]; rw [Nat.cast_mul] · simp only [hP, ↓reduceIte] )]- Project
- ArkLib
- License
- Apache-2.0
- Commit
- fad5cbf80877
- Source
- ArkLib/Data/Probability/Instances.lean:236-295
Reuse this declaration
Bring the exact result into your workflow
The import identifies the source module. Your project still needs the pinned package dependency shown on this page.
What this badge means
This completion status comes from the project or community source. It has not yet been represented here as an independent rebuild and axiom audit.
Continue in this project
Related declarations
Affine gaps lifted to interleaved codes
affine_gaps_lifted_to_interleaved_codes
Project documentation
This lemma proves the final algebraic step in the DG25 Theorem 3.1 proof. It shows that if R > e + 1, then e * (R / (R - 1)) < e + 1. The intuition is that the fraction R / (R - 1) is always greater than 1, but as R gets larger, it gets closer to 1. The hypothesis R > e + 1 provides a strong enough bound to ensure the product e * (fraction) do...
Source project: ArkLib
Person-level attribution pending.
Gadget Decompose coeff
ArkLib.Lattices.Ajtai.gadgetDecompose_coeff
Plain-language statement
The k-th coefficient (k < deg φ) of a gadget-decomposition block is exactly the corresponding digit of the corresponding input coefficient.
Source project: ArkLib
Person-level attribution pending.
Gadget Decompose lawful
ArkLib.Lattices.Ajtai.gadgetDecompose_lawful
Plain-language statement
The base-b gadget decomposition is a lawful gadget decomposition.
Source project: ArkLib
Person-level attribution pending.