Theorem 13, p.44 — balanced randomized affine regret rate for the normalized family
OpenDeepQOperator.Supplement.theorem_13adaptivitybellman-operatorinformation-based-complexityp2o-gran-per-chapterp2o-plan-paperp2o-v1query-complexity
For the normalized family (76)–(78), , , every integer satisfies
The upper-to-lower ratio is at most and tends to as . Thus the rate is for every missing-measurement count. The upper bound uses raw queries. For the value is zero. The exact randomized decision complexity is fixed versus adaptive, for both the affine and the raw oracles.
The theorem pins the randomized fixed affine regret within a constant factor at every budget.
Differences from the source:
- encoding: worst-input expected regret → an -valued supremum of lower Lebesgue integrals of the loss; reason: Lean convention; the loss is nonnegative and bounded, so the values agree
- encoding: at most queries (reply bits) on every input and random path → the algorithm is run with fuel (protocols: hard budget
HardBudgetOnand fuel ); reason: a run with fuel asks at most queries, and an algorithm with hard budget behaves identically under fuel - encoding: a fixed design of at most affine measurements → an
ExactDesignwith exactly rows, zero rows allowed; reason: a zero row is a known constant, so padding to rows changes nothing - encoding: randomized algorithm with private seed; deterministic algorithm → seed space a standard Borel probability space ; deterministic = seed space
Unitwith the Dirac measure; reason: the standing randomization model (20), §5, p.8 - encoding: the minimax value is at least → every algorithm of the class (measurable, on every standard Borel probability seed space) has worst regret at least ; reason: an infimum is at least exactly when every member is
- reading: the minimax value is at most (an upper bound on an infimum) → there is an algorithm of the class whose worst regret is at most ; reason: the source proves each upper bound by an explicit algorithm or design it describes
- reading: for the value is zero → a measurable fixed raw design with rows has zero worst regret; reason: reading all coordinates (raw sites) decides exactly
- reading: the rate is → absolute chosen before with a lower bound for every design and an attaining design; reason: the constants of (80) are absolute
- reading: exact randomized complexity fixed versus adaptive, affine and raw → every measurable design with rows has positive worst regret; a deterministic two-query raw algorithm is exact; every measurable randomized one-query affine algorithm has positive worst regret; reason: exact complexity is two-sided; raw queries are affine
Preamble
import Mathlib import Definitions.Def_DeepQOperator_Family import Definitions.Def_DeepQOperator_ExactDesign import Definitions.Def_DeepQOperator_Supplement_AdaptiveAlg import Definitions.Def_DeepQOperator_Supplement_Normalized open MeasureTheory Filter Topology
Formal statement
namespace DeepQOperator.Supplement
/-- **Theorem 13 (Balanced randomized affine regret rate)** (Q. Qi, *Deep Q-Operator: Fixed-Accuracy
Adaptive Information and Shared Bellman Decisions*, STOC 2027 submission (manuscript sha256
9f8f3285…), Theorem 13, p.44, PDF p.47; proved in §L.2 p.45 and Appendix K).
For the normalized family (76)–(78), `m ≥ 2`, `0 < γ < 1`, and every integer `0 ≤ N ≤ m`,
`L_{m,N} := γ(m+1-N)/(16π²m(√m+1)²) ≤ 𝔇^{fixed,aff,ran}_N ≤ γ min{m, m+1-N}/(4m²)` (80). The
upper-to-lower ratio is at most `4π²(1 + 1/√m)² < 116`, and this bound tends to `4π²` as `m → ∞`.
Thus the rate is `Θ(γ(m+1-N)/m²)` for every missing-measurement count. The upper bound uses raw
queries. For `N ≥ m + 1` the value is zero. The exact randomized decision complexity is `m + 1` for
fixed and `2` for adaptive information, for both the affine and the raw oracle.
**Formalization Note.** `𝔇^{fixed,aff,ran}_N` is the infimum, over measurable randomized fixed designs
`ExactDesign (NCoord m) Ω N` on a standard Borel probability seed space, of the worst expected loss
`designRegret` on `P_m`; the lower bound is stated for every such design, the upper bound by a raw
design attaining it. Randomized adaptive algorithms are measurable `AdaptiveAlg` with exact replies run
with fuel equal to the budget; the deterministic two-query algorithm is the one-point-seed case
(`Unit`, Dirac measure). The `Θ` constants are absolute: chosen before `m`, `γ`, `N`. The ratio is
computed with `γ` cancelled. -/
theorem theorem_13 :
(∀ m : ℕ, 2 ≤ m → ∀ γ : ℝ, 0 < γ → γ < 1 →
(∀ N : ℕ, N ≤ m →
(∀ (Ω : Type) [MeasurableSpace Ω] [StandardBorelSpace Ω] (μ : Measure Ω)
[IsProbabilityMeasure μ] (D : ExactDesign (NCoord m) Ω N), D.IsMeasurable →
ENNReal.ofReal
(γ * ((m : ℝ) + 1 - N) / (16 * Real.pi ^ 2 * m * (Real.sqrt m + 1) ^ 2)) ≤
designRegret (nBox m) (nLoss m γ) μ D) ∧
(∃ (Ω : Type) (_ : MeasurableSpace Ω) (_ : StandardBorelSpace Ω) (μ : Measure Ω)
(_ : IsProbabilityMeasure μ) (D : ExactDesign (NCoord m) Ω N),
D.IsMeasurable ∧ IsRawDesign (nDictionary m) D ∧
designRegret (nBox m) (nLoss m γ) μ D ≤
ENNReal.ofReal (γ * min (m : ℝ) ((m : ℝ) + 1 - N) / (4 * (m : ℝ) ^ 2)))) ∧
-- the value is zero for `N ≥ m + 1`, already for raw designs
(∀ N : ℕ, m + 1 ≤ N →
∃ (Ω : Type) (_ : MeasurableSpace Ω) (_ : StandardBorelSpace Ω) (μ : Measure Ω)
(_ : IsProbabilityMeasure μ) (D : ExactDesign (NCoord m) Ω N),
D.IsMeasurable ∧ IsRawDesign (nDictionary m) D ∧
designRegret (nBox m) (nLoss m γ) μ D = 0) ∧
-- exact randomized fixed complexity `m + 1`: no design with at most `m` rows is exact
(∀ N : ℕ, N ≤ m → ∀ (Ω : Type) [MeasurableSpace Ω] [StandardBorelSpace Ω] (μ : Measure Ω)
[IsProbabilityMeasure μ] (D : ExactDesign (NCoord m) Ω N), D.IsMeasurable →
0 < designRegret (nBox m) (nLoss m γ) μ D) ∧
-- exact randomized adaptive complexity `2`, affine and raw
(∃ A : AdaptiveAlg (NCoord m) Unit, A.IsExact ∧ A.IsRaw (nDictionary m) ∧ A.IsMeasurable ∧
algRegret (nBox m) (nLoss m γ) (Measure.dirac ()) A 2 = 0) ∧
(∀ (Ω : Type) [MeasurableSpace Ω] [StandardBorelSpace Ω] (μ : Measure Ω)
[IsProbabilityMeasure μ] (A : AdaptiveAlg (NCoord m) Ω), A.IsExact → A.IsMeasurable →
0 < algRegret (nBox m) (nLoss m γ) μ A 1)) ∧
-- the upper-to-lower ratio
(∀ m : ℕ, 2 ≤ m → ∀ N : ℕ, N ≤ m →
(min (m : ℝ) ((m : ℝ) + 1 - N) / (4 * (m : ℝ) ^ 2)) /
(((m : ℝ) + 1 - N) / (16 * Real.pi ^ 2 * m * (Real.sqrt m + 1) ^ 2)) ≤
4 * Real.pi ^ 2 * (1 + 1 / Real.sqrt m) ^ 2) ∧
(∀ m : ℕ, 2 ≤ m → 4 * Real.pi ^ 2 * (1 + 1 / Real.sqrt m) ^ 2 < 116) ∧
Tendsto (fun m : ℕ => 4 * Real.pi ^ 2 * (1 + 1 / Real.sqrt m) ^ 2) atTop
(𝓝 (4 * Real.pi ^ 2)) ∧
-- the rate `Θ(γ(m+1-N)/m²)` with absolute constants
(∃ c C : ℝ, 0 < c ∧ 0 < C ∧ ∀ m : ℕ, 2 ≤ m → ∀ γ : ℝ, 0 < γ → γ < 1 → ∀ N : ℕ, N ≤ m →
(∀ (Ω : Type) [MeasurableSpace Ω] [StandardBorelSpace Ω] (μ : Measure Ω)
[IsProbabilityMeasure μ] (D : ExactDesign (NCoord m) Ω N), D.IsMeasurable →
ENNReal.ofReal (c * (γ * ((m : ℝ) + 1 - N) / (m : ℝ) ^ 2)) ≤
designRegret (nBox m) (nLoss m γ) μ D) ∧
(∃ (Ω : Type) (_ : MeasurableSpace Ω) (_ : StandardBorelSpace Ω) (μ : Measure Ω)
(_ : IsProbabilityMeasure μ) (D : ExactDesign (NCoord m) Ω N),
D.IsMeasurable ∧
designRegret (nBox m) (nLoss m γ) μ D ≤
ENNReal.ofReal (C * (γ * ((m : ℝ) + 1 - N) / (m : ℝ) ^ 2)))) := by sorry
end DeepQOperator.Supplement
Source
Qi, Deep Q-Operator: Fixed-Accuracy Adaptive Information and Shared Bellman Decisions, STOC 2027 submission (manuscript sha256 9f8f3285…), Theorem 13 and eq. (80), p.44 (PDF p.47); proof in Appendix K