Repository navigation
feat: the Wald inequality at the bipolar rate and the lower bound on expected stopping times - #441
RemyDegenne wants to merge 15 commits into
Conversation
…expected stopping times
For i.i.d. observations, an e-process E for the i.i.d. null of S₁ and a stopping time τ with
values in ℕ satisfy E_Q[log E_τ] ≤ E_Q[τ + 1] * bipolarRate Q₁ S₁ in EReal, with no
integrability assumption on τ. Its case of the e-process ε + (1 - ε) α⁻¹ 𝟙{τ ≤ n} gives the lower
bound log (1/α) / bipolarRate Q₁ S₁ ≤ E_Q[τ + 1] for every sequential test τ of level α.
EValues roadmap, Layer 6, item 3 (expected stopping times), first bullet.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0196ftMVbgjSJzPQEwdek3kM
AI review — changes requestedEach rubric is judged independently by multiple review agents; the PR merges only once every rubric is green — any rubric that is not green (changes requested, blocked, errored, stale, or not yet run) blocks the merge. See the rubrics.
♻️ = approved on an earlier commit, re-run before merge. Reviewing this diff at head |
… its file Move the facts on the shift under `Measure.infinitePi` to `Probability/ProductMeasure.lean`, the `ℕ`-valued stopping-time lemmas to `Probability/Process/Stopping.lean` and monotone convergence for truncations under `∫ᵉ` to the new `MeasureTheory/Integral/EReal.lean`, all public. Use Mathlib's `EReal.le_of_forall_lt_iff_le` and `Finset.sum_ite_eq` instead of re-deriving them. Factor the repeated proof steps into lemmas and split the long proofs. Cite [EVR26, Note 01] and the paper project the proofs are adapted from. Move the restarted rounds and the Wald inequality for a round to `EValues/BipolarRate/Round.lean`, keeping each new file under 1000 lines. Co-Authored-By: Claude Code <noreply@github.com>
…ping-times # Conflicts: # LMLExplore/Probability/ProductMeasure.lean
- Delete the thin wrapper measurable_comp_shift_iterate and use measurable_apply_of_countable at its uses. - Use IsStoppingTime.measurable_nat in Adapted.measurable_apply_stoppingTime. - Add Measure.infinitePi_of_not_isProbabilityMeasure instead of unfolding Measure.infinitePi. - Add integrable_truncLog_comp and a private integrable_natCast_of_le to remove the repeated bounded-integrability arguments. - Add truncLog_def and the simp lemmas truncLog_zero and truncLog_top. - Drop an unused import and make the shift import private in Round.lean. Co-Authored-By: Claude Code <noreply@github.com>
…ping-times # Conflicts: # LMLExplore/Probability/ProductMeasure.lean
abs_truncLog_le, integrable_truncLog_comp and the Wald inequality for a round integral_truncLog_le_bipolarRate_mul no longer assume log δ ≤ c, via the new lower bound ENNReal.min_log_le_truncLog. Co-Authored-By: Claude Code <noreply@github.com>
Co-Authored-By: Claude Code <noreply@github.com>
|
Do check what the conflict actually is before resolving it mechanically. If |
…ping-times # Conflicts: # LMLExplore/Probability/ProductMeasure.lean
|
Merged
Local checks (not CI): |
Make eintegral_min_log_le_integral_truncLog_sub public in LMLExplore/MeasureTheory/Integral/EReal.lean, delete its private copy in Universal.lean, and use it in the Wald inequality of StoppingTime.lean instead of the inline re-derivation. Co-Authored-By: Claude Code <noreply@github.com>
This pull request proves the Wald inequality at the bipolar rate for the e-processes of an i.i.d. null, and its consequence, the lower bound
log (1/α) / don the expected number of observations of a sequential test of levelα, whered = bipolarRate Q₁ S₁. It adds two new files inLMLExplore/EValues/BipolarRate/:Round.leanholds the restarted rounds and the Wald inequality for a round, andStoppingTime.leanholds the Wald inequality for e-processes and the lower bound. It also addsLMLExplore/Analysis/SpecialFunctions/Log/TruncLog.lean,LMLExplore/MeasureTheory/Integral/EReal.leanandLMLExplore/Logic/Function/DependsOn.lean, and public general lemmas in three existing files (listed below). No existing declaration is removed or renamed.Main declarations
Write
Q = Measure.infinitePi fun _ ↦ Q₁. The observations are the coordinates ofℕ → 𝓧withFiltration.piLE, and the null is(fun P ↦ Measure.infinitePi fun _ ↦ P) '' S₁with∀ P ∈ S₁, IsProbabilityMeasure P, as inLMLExplore/EValues/BipolarRate/EProcess.lean.IsEProcess.eintegral_log_stoppedValue_le_mul_bipolarRate: for an e-processEfor the i.i.d. null and a stopping timeτ : (ℕ → 𝓧) → ℕ,∫ᵉ ω, ENNReal.log (E (τ ω) ω) ∂Q ≤ ((∫⁻ ω, (τ ω : ℝ≥0∞) + 1 ∂Q : ℝ≥0∞) : EReal) * bipolarRate Q₁ S₁.No assumption on
Q₁and none on the expectation ofτ: when it is infinite the right side is⊤, or0ifd = 0. Withτconstant, this isIsEProcess.eintegral_log_le_mul_bipolarRateof feat: e-processes for an i.i.d. null grow at most at the bipolar rate in expectation [RR26, Lemma 3.3] #413.toENNReal_log_inv_div_bipolarRate_le_lintegral: for a probability measureQ₁, a stopping timeτ : (ℕ → 𝓧) → WithTop ℕandα : ℝ≥0∞withMeasure.infinitePi (fun _ ↦ P) {ω | τ ω ≠ ⊤} ≤ αfor everyP ∈ S₁,(ENNReal.log α⁻¹).toENNReal / (bipolarRate Q₁ S₁).toENNReal ≤ ∫⁻ ω, ENat.toENNReal (τ ω) + 1 ∂Q.For
d = 0andα < 1, and forα = 0andd < ∞, the expected number of observations is infinite.Other public declarations:
LMLExplore/EValues/BipolarRate/Round.lean:DependsOnBefore σ f(withdependsOnBefore_iff):fis determined by the firstσobservations.integral_truncLog_le_bipolarRate_mul, the Wald inequality for a round.LMLExplore/Analysis/SpecialFunctions/Log/TruncLog.lean(new):ENNReal.truncLog c x, the logarithm ofx : ℝ≥0∞truncated above atc, as a real number, withENNReal.truncLog_def,ENNReal.measurable_truncLog,ENNReal.truncLog_of_ne,ENNReal.truncLog_topandENNReal.truncLog_zero(both@[simp]),ENNReal.integrable_truncLog_comp,ENNReal.truncLog_le,ENNReal.min_log_le_truncLog,ENNReal.log_le_truncLog,ENNReal.abs_truncLog_le,ENNReal.ofReal_exp_truncLog_leandENNReal.min_log_le_truncLog_sub.LMLExplore/Probability/ProductMeasure.lean:MeasureTheory.Measure.infinitePi_of_not_isProbabilityMeasure(the infinite product is0unless every factor is a probability measure),MeasureTheory.shift_iterate_eqOn(sequences that agree on the firsttterms have shifts bysthat agree on the firstt - s),MeasureTheory.measurePreserving_shift_iterate,indepFun_initSeg_shift_iterate,lintegral_mul_comp_shift_iterateandintegral_mul_comp_shift_iterate, about the shift(shift 𝓧)^[t]underMeasure.infinitePi.LMLExplore/Probability/Process/Stopping.lean:MeasureTheory.IsStoppingTime.measurable_nat,IsStoppingTime.natCast_min,IsStoppingTime.toNat_min(with the cast lemmaENat.natCast_toNat_min) andIsStoppingTime.eq_of_forall_le(the last one forFiltration.piLE).LMLExplore/Logic/Function/DependsOn.lean:dependsOn_indicator_of_eqOn, for any Pi type and coordinate sets: if membership inAandfonAare determined by the coordinates ins, so isA.indicator f.LMLExplore/MeasureTheory/Integral/EReal.lean:MeasureTheory.eintegral_le_of_forall_min_le, monotone convergence for the truncationsmin f kunder∫ᵉ, andMeasureTheory.eintegral_min_log_le_integral_truncLog_sub, mixing with a constant in expectation:∫ᵉ min (log E) k ≤ ∫ truncLog k ((1 - ε) E + ε) - log (1 - ε)under a probability measure. The latter was a private lemma ofLMLExplore/EValues/BipolarRate/Universal.lean, whose private copy is deleted; it is now used there and inStoppingTime.lean.LMLExplore/EValues/BipolarRate/Basic.lean:ProbabilityTheory.coe_toReal_bipolarRate.Roadmap coverage (EValues, Layer 6, item 3)
Covered: from the first bullet ("Lower bound"), the lower bound
𝔼_Q[τ + 1] ≥ log (1/α) / dfor every level-αsequential test, for anyS₁andQ₁, with an infinite expectation whend = 0, and the Wald inequality𝔼_Q[log E_τ] ≤ d 𝔼_Q[τ + 1]inERealfor every e-process and every stopping time with values inℕ, with Statlib's∫ᵉon the left. The lower bound is derived from the Wald inequality, applied to the e-processε + (1 - ε) α⁻¹ 𝟙{τ ≤ n}at the stopping timesmin τ k, as the roadmap describes. The form withKL_inf(Q₁, S₁)in place ofdfollows frombipolarRate_le_iInf_klDivand is not stated separately.Left: the almost sure form
liminf (τ (exp (-n)) / n) ≥ 1 / KL_inf(Q₁, S₁)of the same bullet [ARR26, Theorem 3.1]; the second bullet (tests attaining the bound withdto first order); first passage times; the upper bound [ARR26, Theorem 4.1]; regret and follow-the-leader bets. Row R10 of the paper project does not contain them. The other half of row R10, the almost sure boundlimsup (1/n) log (M n) ≤ dof Layer 6, item 1, is ported in a separate pull request.Notes for reviewers
σ ≥ 1, bounded, and a payoffV ≥ δ > 0, both determined by the observations of the round, with expectation at most1on each event{σ ≤ n}under the i.i.d. null. The round is restarted on the fresh observations each time it ends (Tau Ceti'sTauCeti.Probability.shift, iterated). Within the firstNobservations, the geometric mixture overjof the products of the payoffs of the firstjrounds, on the event that they end by timeN, is anN-sample e-variable: by induction onj, splitting along the valuetof the restart time, using that under an i.i.d. law the firsttobservations are independent of(shift 𝓧)^[t]. Its expected logarithm underQis at mostmaxLogGrowth Q₁ S₁ N ≤ N d(eintegral_log_le_maxLogGrowth_of_dependsOn). Bounding the logarithm of the mixture below by its term at the number of completed rounds, and a Wald argument (the rounds that start within the firstNobservations use more thanNof them), give𝔼[truncated log V] ≤ d 𝔼[σ]asN → ∞. For the Wald inequality, the round is the e-process atmin τ kmixed with a constant,(1 - ε) E_{min τ k} + ε; thenk → ∞(dominated convergence), the truncation level→ ∞(monotone convergence for∫ᵉ) andε → 0. The proof needs neither the law of large numbers nor the strong Markov property, and the cased = 0with an infinite expectation ofτneeds no separate argument, since the bound formin τ kis0for everyk.[EVR26, Note 01], and the bounds of [ARR26] and [WSJ26] haveKL_inf(Q₁, S₁), which can be larger [RR26, Proposition 5.4]. The proofs are rewrites of the Lean proofs of the paper project (EValuesResearch/Results/BipolarRate.lean,BipolarRateAux.lean), specialized to stopping times with values inℕand bounded rounds, which removes the infinite restart times and the separate argument ford = 0of the source.𝔼[τ + 1] * d, notd 𝔼[τ + 1], asIsEProcess.eintegral_log_le_mul_bipolarRatewrites(n + 1) * d.ℝ≥0∞, withlog (1/α)as(ENNReal.log α⁻¹).toENNReal(so0forα ≥ 1and∞forα = 0) anddas(bipolarRate Q₁ S₁).toENNReal. The product formlog (1/α) ≤ d * 𝔼[τ + 1]inERealwould be false ford = 0, where𝔼[τ + 1] = ∞and0 * ⊤ = 0.P^ℕ {τ ≠ ⊤} ≤ α, the roadmap'sP (τ < ∞) ≤ αwith LMLExplore's≠ ⊤. The expected number of observations of aWithTop ℕ-valuedτis∫⁻ ENat.toENNReal (τ ω) + 1, sinceWithTop ℕhas no coercion toℝ≥0∞.Q₁: ifQ₁is not a probability measure, its i.i.d. law is0and both sides are0. The lower bound needsIsProbabilityMeasure Q₁. Both need the members ofS₁to be probability measures: the e-process condition is about the lawsP^ℕ, anddabout the lawsP^n.ℕ-valued stopping times and about the∫ᵉtruncations are public, in the generic files listed above.sum_indicator_eq_indicator_lestays private: it is Mathlib'sFinset.sum_ite_eqafter unfoldingSet.indicator.Measure.infinitePiinLMLExplore/Probability/ProductMeasure.lean.Roadmap: EValues
Paper: RemyDegenne/evalues_research
Item: R10
🤖 Generated with Claude Code
https://claude.ai/code/session_0196ftMVbgjSJzPQEwdek3kM