Theorem 24, p.58 — randomized fixed affine information with a truncated-Gaussian prior
OpenDeepQOperator.Supplement.theorem_24adaptivitybellman-operatorinformation-based-complexityp2o-gran-per-chapterp2o-plan-paperp2o-v1query-complexity
For the normalized family (76)–(78), , , set
(natural logarithm). Every randomized fixed design of at most arbitrary affine measurements satisfies
For the matching-scale upper bound
is attained by a fixed randomized raw-query design. The exact randomized decision complexity is for fixed affine information and for adaptive affine or raw information. The lower bound allows arbitrary real coefficients and measurable randomized decoders, with no bound on their processing time.
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: exact randomized complexity fixed affine, adaptive affine or raw → a design with rows is exact and every measurable design with rows has positive worst regret; a deterministic two-query raw algorithm is exact and every measurable one-query affine algorithm has positive worst regret; reason: exact complexity is two-sided
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
Formal statement
namespace DeepQOperator.Supplement
/-- **Theorem 24 (Randomized fixed affine information)** (Q. Qi, *Deep Q-Operator: Fixed-Accuracy
Adaptive Information and Shared Bellman Decisions*, STOC 2027 submission (manuscript sha256
9f8f3285…), Theorem 24, p.58, PDF p.61; proved in Appendix O).
For the normalized family (76)–(78), `m ≥ 2`, `0 < γ < 1`, put `L_m = √(2 log(2^16 (m+1) m²))`
(natural logarithm). Every randomized fixed design of at most `N ≤ m` arbitrary affine measurements
satisfies `𝔇^{fixed,aff,ran}_N ≥ γ(m+1-N)/(512 m² L_m)` (102). For `2 ≤ N ≤ m` the matching-scale upper
bound is `𝔇^{fixed,aff,ran}_N ≤ γ(m+1-N)/(4m²)` (103), attained by a fixed randomized raw-query design.
The exact randomized decision complexity is `m + 1` for fixed affine information and `2` for adaptive
affine or raw information. The lower bound allows arbitrary real coefficients and measurable randomized
decoders, with no bound on their processing time.
**Formalization Note.** Fixed designs are measurable `ExactDesign (NCoord m) Ω N` (arbitrary real rows
and offsets, jointly measurable decoder) on a standard Borel probability seed space; "at most `N`" is
`N` rows, zero rows allowed. The upper bound is a raw design attaining (103). 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. `log` is `Real.log`. -/
theorem theorem_24 :
∀ 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) /
(512 * (m : ℝ) ^ 2 *
Real.sqrt (2 * Real.log ((2 : ℝ) ^ 16 * ((m : ℝ) + 1) * (m : ℝ) ^ 2)))) ≤
designRegret (nBox m) (nLoss m γ) μ D) ∧
(∀ N : ℕ, 2 ≤ N → N ≤ m →
∃ (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 * ((m : ℝ) + 1 - N) / (4 * (m : ℝ) ^ 2))) ∧
-- exact randomized fixed affine complexity `m + 1`: a design with `m + 1` rows is exact
(∃ (Ω : Type) (_ : MeasurableSpace Ω) (_ : StandardBorelSpace Ω) (μ : Measure Ω)
(_ : IsProbabilityMeasure μ) (D : ExactDesign (NCoord m) Ω (m + 1)),
D.IsMeasurable ∧ designRegret (nBox m) (nLoss m γ) μ D = 0) ∧
-- and 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) := 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 24 and eqs. (102)–(103), p.58 (PDF p.61); proof in Appendix O