AFTD Auto-Formalizing Theoretical Domains

dp_suffix_val

def

The dynamic programming value v_i(r) for the suffix from item i in {0, ..., n} under subsampling rate r, satisfying v_i(r) = v_{i+1}(r) + r * E[(X_i - v_{i+1}(r))^+] for i < n and v_n(r) = 0.

Statement

open MeasureTheory ProbabilityTheory in
noncomputable def dp_suffix_val {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω)
    (n : ℕ) (X : Fin n → Ω → ℝ) (r : ℝ) (i : ℕ) : ℝ
Defined in
arXiv:2611.15711, equation (2)
Verified
14 Nov 2026
Axioms
none listed
Used by 1
  • dp_thresholdGiven a measurable space Ω, a measure P on Ω, a natural number n, a sequence of functions X : Fin n → Ω → ℝ, a real parameter r, and an…

Read back from the Lean

For a measurable space Ω, a measure P on Ω, a natural number n, a sequence of functions X : Fin n → Ω → ℝ, a real parameter r, and an index i : ℕ, dp_suffix_val P n X r i is defined by well-founded recursion on n - i as follows: - If i < n, it returns dp_suffix_val P n X r (i + 1) + r * ∫ ω, max (X ⟨i, h⟩ ω - dp_suffix_val P n X r (i + 1)) 0 ∂P. - If i ≥ n, it returns 0.

Written by a model that saw only the Lean, never the English above. If the two disagree, that is worth a challenge.

Lean source view module on GitHub

open MeasureTheory ProbabilityTheory in
/-- Dynamic programming value function for suffix from item i under subsampling rate r. -/
noncomputable def dp_suffix_val {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω)
    (n : ℕ) (X : Fin n → Ω → ℝ) (r : ℝ) (i : ℕ) : ℝ :=
  if h : i < n then
    dp_suffix_val P n X r (i + 1) + r * ∫ ω, max (X ⟨i, h⟩ ω - dp_suffix_val P n X r (i + 1)) 0 ∂P
  else
    0
termination_by n - i
decreasing_by omega
See it in the graphDiscuss this resultChallenge it