Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proposition 4 — TPG under general Bernoulli randomization is unbiased for the truncated policy gradient

Proved
TPG.proposition_4_general_bernoulli

by Wenqian Xing · Oct 12, 2026 · Mathlib 0df444a (Lean v4.33.1)

causal-inferencemarkov-chainnonstationaritypolicy-gradienttreatment-effect

Setting. SSS is a finite state space and the action Zt∈{0,1}Z_t\in\{0,1\}Zt​∈{0,1} is control or treatment. Time is t=0,1,…,T−1t=0,1,\dots,T-1t=0,1,…,T−1, which is the source's time t+1t+1t+1. Under action zzz at time ttt the state moves according to a transition matrix PztP_z^tPzt​, i.e. P(Xt+1=x′∣Xt=x,Zt=z)=Pzt(x,x′)\mathbb P(X_{t+1}=x'\mid X_t=x,Z_t=z)=P_z^t(x,x')P(Xt+1​=x′∣Xt​=x,Zt​=z)=Pzt​(x,x′); the matrices may depend on ttt arbitrarily. The initial state X0X_0X0​ has distribution ρ\rhoρ, and the reward YtY_tYt​ has distribution R(⋅∣Xt,Zt)R(\cdot\mid X_t,Z_t)R(⋅∣Xt​,Zt​) on R\mathbb RR, with R([−M,M]∣x,z)=1R([-M,M]\mid x,z)=1R([−M,M]∣x,z)=1 for all x,zx,zx,z and conditional independence from everything else given (Xt,Zt)(X_t,Z_t)(Xt​,Zt​). Under the policy πθ\pi_\thetaπθ​ every action is treatment with probability θ\thetaθ, independently of everything else; Eθ\mathbb E_\thetaEθ​, Pθ\mathbb P_\thetaPθ​ and Var⁡θ\operatorname{Var}_\thetaVarθ​ refer to the law of the trajectory (Xt,Zt,Yt)t<T(X_t,Z_t,Y_t)_{t<T}(Xt​,Zt​,Yt​)t<T​ under πθ\pi_\thetaπθ​.

Fix an experimental allocation θ∈(0,1)\theta\in(0,1)θ∈(0,1) and a truncation size kkk. With the score sθ(z)=zθ−1−z1−θs_\theta(z)=\dfrac{z}{\theta}-\dfrac{1-z}{1-\theta}sθ​(z)=θz​−1−θ1−z​, the TPG estimator and the truncated QQQ-values under πθ\pi_\thetaπθ​ are

τ^θ,kTPG=1T∑u=0T−1sθ(Zu)∑t=umin⁡(u+k, T−1)Yt,Qθu,k(z)=∑t=umin⁡(u+k, T−1)Eθ[Yt∣Zu=z].\widehat\tau^{\mathrm{TPG}}_{\theta,k}=\frac1T\sum_{u=0}^{T-1}s_\theta(Z_u)\sum_{t=u}^{\min(u+k,\,T-1)}Y_t,\qquad Q_\theta^{u,k}(z)=\sum_{t=u}^{\min(u+k,\,T-1)}\mathbb E_\theta[Y_t\mid Z_u=z].τθ,kTPG​=T1​u=0∑T−1​sθ​(Zu​)t=u∑min(u+k,T−1)​Yt​,Qθu,k​(z)=t=u∑min(u+k,T−1)​Eθ​[Yt​∣Zu​=z].

Write Pqt=qP1t+(1−q)P0tP_q^t=qP_1^t+(1-q)P_0^tPqt​=qP1t​+(1−q)P0t​, Pqa→b=Pqa⋯Pqb−1P_q^{a\to b}=P_q^{a}\cdots P_q^{b-1}Pqa→b​=Pqa​⋯Pqb−1​, rq(x)=q r(x,1)+(1−q) r(x,0)r_q(x)=q\,r(x,1)+(1-q)\,r(x,0)rq​(x)=qr(x,1)+(1−q)r(x,0) with r(x,z)=∫y R(dy∣x,z)r(x,z)=\int y\,R(dy\mid x,z)r(x,z)=∫yR(dy∣x,z), and μqt=ρ⊤Pq0→t\mu_q^t=\rho^\top P_q^{0\to t}μqt​=ρ⊤Pq0→t​. The hypothetical truncated policy value is, with at=max⁡(0,t−k)a_t=\max(0,t-k)at​=max(0,t−k),

Jk,θ(q)=1T∑t<Tμθat Pqat→t rq,J_{k,\theta}(q)=\frac1T\sum_{t<T}\mu_\theta^{a_t}\,P_q^{a_t\to t}\,r_q ,Jk,θ​(q)=T1​t<T∑​μθat​​Pqat​→t​rq​,

and J(q)=1T∑t<TEq[Yt]J(q)=\frac1T\sum_{t<T}\mathbb E_q[Y_t]J(q)=T1​∑t<T​Eq​[Yt​] is the value function.

Proposition. For every T≥1T\ge1T≥1, 0≤k≤T0\le k\le T0≤k≤T and θ∈(0,1)\theta\in(0,1)θ∈(0,1):

  1. Eθ[τ^θ,kTPG]=1T∑u=0T−1[Qθu,k(1)−Qθu,k(0)]\displaystyle \mathbb E_\theta\bigl[\widehat\tau^{\mathrm{TPG}}_{\theta,k}\bigr]=\frac1T\sum_{u=0}^{T-1}\bigl[Q_\theta^{u,k}(1)-Q_\theta^{u,k}(0)\bigr]Eθ​[τθ,kTPG​]=T1​u=0∑T−1​[Qθu,k​(1)−Qθu,k​(0)];
  2. q↦Jk,θ(q)q\mapsto J_{k,\theta}(q)q↦Jk,θ​(q) is differentiable at q=θq=\thetaq=θ with derivative equal to the same quantity, ∂∂qJk,θ(q)∣q=θ=1T∑u=0T−1[Qθu,k(1)−Qθu,k(0)]\displaystyle\frac{\partial}{\partial q}J_{k,\theta}(q)\Big|_{q=\theta}=\frac1T\sum_{u=0}^{T-1}\bigl[Q_\theta^{u,k}(1)-Q_\theta^{u,k}(0)\bigr]∂q∂​Jk,θ​(q)​q=θ​=T1​u=0∑T−1​[Qθu,k​(1)−Qθu,k​(0)];
  3. for k=Tk=Tk=T, JT,θ(q)=J(q)J_{T,\theta}(q)=J(q)JT,θ​(q)=J(q) for all q∈[0,1]q\in[0,1]q∈[0,1];
  4. JJJ is differentiable at θ\thetaθ and Eθ[τ^θ,TTPG]=J′(θ)\mathbb E_\theta\bigl[\widehat\tau^{\mathrm{TPG}}_{\theta,T}\bigr]=J'(\theta)Eθ​[τθ,TTPG​]=J′(θ).

This extends the policy-gradient interpretation of the main text (Propositions 1 and 3) from θ=1/2\theta=1/2θ=1/2 to every interior experimental allocation.

Formalization Note Jk,θJ_{k,\theta}Jk,θ​ is defined by the matrix expression of the source, with t−kt-kt−k a truncated subtraction; item 3 is stated on [0,1][0,1][0,1], where πq\pi_qπq​ is a policy.

Preamble
import Mathlib
import Definitions.Def_TPG_Model
import Definitions.Def_TPG_Estimators

open MeasureTheory ProbabilityTheory
open scoped BigOperators ENNReal
Formal statement
namespace TPG
theorem proposition_4_general_bernoulli {S : Type} [Fintype S] [DecidableEq S] [MeasurableSpace S] [MeasurableSingletonClass S]
    (E : Env S) (T k : ℕ) (θ M : ℝ)
    (hT : 0 < T) (hk : k ≤ T) (hθ0 : 0 < θ) (hθ1 : θ < 1) (hM : E.RewardBound M) :
    (∫ ω, tpgEst θ T k ω ∂(E.expLaw T θ)
        = (1 / (T : ℝ)) * ∑ u : Fin T, (E.Qk θ T k u true - E.Qk θ T k u false))
      ∧ HasDerivAt (fun q : ℝ => E.JkGen T k θ q)
          ((1 / (T : ℝ)) * ∑ u : Fin T, (E.Qk θ T k u true - E.Qk θ T k u false)) θ
      ∧ (∀ q : ℝ, 0 ≤ q → q ≤ 1 → E.JkGen T T θ q = E.J T q)
      ∧ HasDerivAt (fun q : ℝ => E.J T q) (∫ ω, tpgEst θ T T ω ∂(E.expLaw T θ)) θ := by sorry
end TPG
Source
Johari, Peng, Xing, Estimation of Treatment Effects Under Nonstationarity via the Truncated Policy Gradient Estimator, revised manuscript of October 2026 (revision of arXiv:2506.05308v3, https://arxiv.org/abs/2506.05308); numbering follows the revised manuscript. Appendix D, pp. 31–32, Proposition 4, eqs. (11)–(12); proof in Appendix G.4, p. 46.
Read-back

What the Lean code literally says, in plain math · claude-opus-5-5

What is quantified. The statement holds for every choice of the following data.

  • A finite set SSS, carrying an arbitrary σ\sigmaσ-algebra in which every singleton is measurable (as SSS is finite, every subset of SSS is then measurable). Equality of elements of SSS is assumed decidable, a technical assumption with no mathematical content (it is used to form identity matrices).

  • An "environment" EEE on SSS, meaning the following package:

    • for every natural number t∈Nt \in \mathbb{N}t∈N (all of them, not only those below the horizon) and every Boolean value z∈{true,false}z \in \lbrace \mathrm{true}, \mathrm{false} \rbracez∈{true,false}, a real S×SS \times SS×S matrix PtzP_t^{z}Ptz​ whose entries are all ≥0\ge 0≥0 and whose rows each sum to 111;
    • a function ρ:S→R\rho : S \to \mathbb{R}ρ:S→R with ρ(x)≥0\rho(x) \ge 0ρ(x)≥0 for all xxx and ∑x∈Sρ(x)=1\sum_{x \in S} \rho(x) = 1∑x∈S​ρ(x)=1;
    • for every x∈Sx \in Sx∈S and every Boolean zzz, a Borel probability measure Rx,zR_{x,z}Rx,z​ on R\mathbb{R}R.

    Because ρ\rhoρ sums to 111, such an EEE can only exist when SSS is nonempty.

  • Natural numbers TTT and kkk, with the hypotheses 0<T0 \lt T0<T and k≤Tk \le Tk≤T. The value k=0k = 0k=0 is allowed.

  • A real number θ\thetaθ with the hypotheses 0<θ0 \lt \theta0<θ and θ<1\theta \lt 1θ<1 (both strict).

  • A real number MMM (no sign condition), with the hypothesis that Rx,z([−M,M])=1R_{x,z}([-M, M]) = 1Rx,z​([−M,M])=1 for every x∈Sx \in Sx∈S and every Boolean zzz. The number MMM occurs nowhere else in the statement. If M<0M \lt 0M<0 the interval [−M,M][-M,M][−M,M] is empty and this hypothesis cannot be satisfied.

The conclusion is the conjunction of four claims, (a), (b), (c), (d) below. The objects they involve are defined first.

The sample space. Let Ω\OmegaΩ be the set of all sequences ω=(ω0,…,ωT−1)\omega = (\omega_0, \dots, \omega_{T-1})ω=(ω0​,…,ωT−1​) with each ωt∈S×{true,false}×R\omega_t \in S \times \lbrace \mathrm{true}, \mathrm{false} \rbrace \times \mathbb{R}ωt​∈S×{true,false}×R, with the product σ\sigmaσ-algebra (the given one on SSS, the discrete one on the Booleans, the Borel one on R\mathbb{R}R). Write Xt(ω)∈SX_t(\omega) \in SXt​(ω)∈S, Zt(ω)∈{true,false}Z_t(\omega) \in \lbrace \mathrm{true}, \mathrm{false} \rbraceZt​(ω)∈{true,false} and Yt(ω)∈RY_t(\omega) \in \mathbb{R}Yt​(ω)∈R for the three components of ωt\omega_tωt​. Times run over 0,1,…,T−10, 1, \dots, T-10,1,…,T−1.

The measures μq\mu_qμq​. Let qqq be any real number (not only one in [0,1][0,1][0,1]). For x=(x0,…,xT−1)∈STx = (x_0,\dots,x_{T-1}) \in S^Tx=(x0​,…,xT−1​)∈ST and a Boolean sequence z=(z0,…,zT−1)z = (z_0,\dots,z_{T-1})z=(z0​,…,zT−1​) define the real number

wq(x,z)=ρ(x0)⋅∏s=0T−2Pszs(xs,xs+1)⋅∏t=0T−1bq(zt),bq(true)=q,bq(false)=1−q.w_q(x,z) = \rho(x_0) \cdot \prod_{s=0}^{T-2} P_s^{z_s}(x_s, x_{s+1}) \cdot \prod_{t=0}^{T-1} b_q(z_t), \qquad b_q(\mathrm{true}) = q, \quad b_q(\mathrm{false}) = 1-q .wq​(x,z)=ρ(x0​)⋅s=0∏T−2​Pszs​​(xs​,xs+1​)⋅t=0∏T−1​bq​(zt​),bq​(true)=q,bq​(false)=1−q.

(The middle product is empty, equal to 111, when T=1T = 1T=1. The matrix PszsP_s^{z_s}Pszs​​ governs the passage from index sss to index s+1s+1s+1.) Then

μq=∑(x,z)max⁡(wq(x,z),0)⋅⨂t=0T−1(δxt⊗δzt⊗Rxt,zt),\mu_q = \sum_{(x,z)} \max\bigl(w_q(x,z), 0\bigr) \cdot \bigotimes_{t=0}^{T-1} \bigl( \delta_{x_t} \otimes \delta_{z_t} \otimes R_{x_t, z_t} \bigr),μq​=(x,z)∑​max(wq​(x,z),0)⋅t=0⨂T−1​(δxt​​⊗δzt​​⊗Rxt​,zt​​),

the sum running over all ∣S∣T⋅2T\lvert S \rvert^T \cdot 2^T∣S∣T⋅2T pairs (x,z)(x,z)(x,z), where δ\deltaδ denotes a point mass and the big tensor product is the product measure on Ω\OmegaΩ. Under the (x,z)(x,z)(x,z)-th summand, Xt=xtX_t = x_tXt​=xt​ and Zt=ztZ_t = z_tZt​=zt​ for all ttt almost surely, and Y0,…,YT−1Y_0,\dots,Y_{T-1}Y0​,…,YT−1​ are independent with YtY_tYt​ distributed according to Rxt,ztR_{x_t,z_t}Rxt​,zt​​.

Negative weights are replaced by 000. When 0≤q≤10 \le q \le 10≤q≤1 no weight is negative, μq\mu_qμq​ is a probability measure, and it is the joint law of the following process: X0X_0X0​ has distribution ρ\rhoρ; each ZtZ_tZt​ is true\mathrm{true}true with probability qqq, independently of X0,Z0,…,Xt−1,Zt−1,XtX_0, Z_0, \dots, X_{t-1}, Z_{t-1}, X_tX0​,Z0​,…,Xt−1​,Zt−1​,Xt​; given the past, Xt+1X_{t+1}Xt+1​ has distribution PtZt(Xt,⋅)P_t^{Z_t}(X_t, \cdot)PtZt​​(Xt​,⋅); and, given all the XtX_tXt​ and ZtZ_tZt​, the YtY_tYt​ are independent with YtY_tYt​ distributed according to RXt,ZtR_{X_t, Z_t}RXt​,Zt​​. When qqq lies outside [0,1][0,1][0,1] some weights are negative and are set to 000, and μq\mu_qμq​ is then a finite measure that is in general not of total mass 111.

The function JJJ. For every real qqq,

J(q)=1T∑t=0T−1∫ΩYt dμq,J(q) = \frac{1}{T} \sum_{t=0}^{T-1} \int_\Omega Y_t \ \mathrm{d}\mu_q ,J(q)=T1​t=0∑T−1​∫Ω​Yt​ dμq​,

a function R→R\mathbb{R} \to \mathbb{R}R→R defined for all real qqq. Integrals are Lebesgue (Bochner) integrals, which by library convention are 000 if the integrand is not integrable; under the hypothesis on MMM each YtY_tYt​ satisfies ∣Yt∣≤M\lvert Y_t \rvert \le M∣Yt​∣≤M almost everywhere for each measure involved.

The random variable GjG_jGj​. For a natural number jjj (used below with j=kj = kj=k and with j=Tj = Tj=T) define on Ω\OmegaΩ

Gj(ω)=1T∑u=0T−1sθ(Zu(ω))⋅∑t : u≤t≤u+j, t≤T−1Yt(ω),sθ(true)=1θ,sθ(false)=−11−θ.G_j(\omega) = \frac{1}{T} \sum_{u=0}^{T-1} s_\theta\bigl(Z_u(\omega)\bigr) \cdot \sum_{t \ : \ u \le t \le u+j, \ t \le T-1} Y_t(\omega), \qquad s_\theta(\mathrm{true}) = \frac{1}{\theta}, \quad s_\theta(\mathrm{false}) = -\frac{1}{1-\theta} .Gj​(ω)=T1​u=0∑T−1​sθ​(Zu​(ω))⋅t : u≤t≤u+j, t≤T−1∑​Yt​(ω),sθ​(true)=θ1​,sθ​(false)=−1−θ1​.

The inner sum is over the times ttt in the window from uuu to u+ju+ju+j inclusive, cut off at the last time T−1T-1T−1. For j=Tj = Tj=T the window is every ttt with u≤t≤T−1u \le t \le T-1u≤t≤T−1.

The numbers Quk,zQ_u^{k,z}Quk,z​ and DkD_kDk​. For a measure μ\muμ on Ω\OmegaΩ and an event AAA, let μ(⋅∣A)\mu(\cdot \mid A)μ(⋅∣A) denote the measure B↦μ(A)−1μ(A∩B)B \mapsto \mu(A)^{-1} \mu(A \cap B)B↦μ(A)−1μ(A∩B) (by library convention this is the zero measure if μ(A)=0\mu(A) = 0μ(A)=0; here it is used only with μ=μθ\mu = \mu_\thetaμ=μθ​ and A={Zu=z}A = \lbrace Z_u = z \rbraceA={Zu​=z}, events of μθ\mu_\thetaμθ​-measure θ\thetaθ and 1−θ1 - \theta1−θ, both positive). For each time u∈{0,…,T−1}u \in \lbrace 0, \dots, T-1 \rbraceu∈{0,…,T−1} and each Boolean zzz define

Quk,z=∑t : u≤t≤u+k, t≤T−1∫ΩYt dμθ(⋅∣Zu=z),Q_u^{k,z} = \sum_{t \ : \ u \le t \le u+k, \ t \le T-1} \int_\Omega Y_t \ \mathrm{d}\mu_\theta(\cdot \mid Z_u = z),Quk,z​=t : u≤t≤u+k, t≤T−1∑​∫Ω​Yt​ dμθ​(⋅∣Zu​=z),

the sum over the same kind of window (with j=kj = kj=k) of the conditional expectation of YtY_tYt​ under μθ\mu_\thetaμθ​ given the single event {Zu=z}\lbrace Z_u = z \rbrace{Zu​=z}. The conditioning is on the value of ZuZ_uZu​ only, not on XuX_uXu​ or anything else. Put

Dk=1T∑u=0T−1(Quk,true−Quk,false).D_k = \frac{1}{T} \sum_{u=0}^{T-1} \bigl( Q_u^{k,\mathrm{true}} - Q_u^{k,\mathrm{false}} \bigr).Dk​=T1​u=0∑T−1​(Quk,true​−Quk,false​).

The matrix expression J~j(θ,q)\widetilde J_j(\theta, q)Jj​(θ,q). This is built from matrices only, with no measure on Ω\OmegaΩ and no replacement of negative numbers by 000. For a Boolean zzz and x∈Sx \in Sx∈S let rˉz(x)=∫Ry dRx,z(y)\bar r^{z}(x) = \int_{\mathbb{R}} y \ \mathrm{d}R_{x,z}(y)rˉz(x)=∫R​y dRx,z​(y) be the mean of Rx,zR_{x,z}Rx,z​. For a real number ppp put

rp(x)=p⋅rˉtrue(x)+(1−p)⋅rˉfalse(x),Pt(p)=p⋅Pttrue+(1−p)⋅Ptfalse,r_p(x) = p \cdot \bar r^{\mathrm{true}}(x) + (1-p) \cdot \bar r^{\mathrm{false}}(x), \qquad P_t^{(p)} = p \cdot P_t^{\mathrm{true}} + (1-p) \cdot P_t^{\mathrm{false}},rp​(x)=p⋅rˉtrue(x)+(1−p)⋅rˉfalse(x),Pt(p)​=p⋅Pttrue​+(1−p)⋅Ptfalse​,

and for natural numbers a,ba, ba,b let Πp(a,b)=Pa(p)Pa+1(p)⋯Pb−1(p)\Pi_p(a,b) = P_a^{(p)} P_{a+1}^{(p)} \cdots P_{b-1}^{(p)}Πp​(a,b)=Pa(p)​Pa+1(p)​⋯Pb−1(p)​ be the ordered matrix product of the b−ab - ab−a matrices with indices a,a+1,…,b−1a, a+1, \dots, b-1a,a+1,…,b−1, understood to be the identity matrix when b≤ab \le ab≤a. For a natural number jjj (used below with j=kj = kj=k and j=Tj = Tj=T), the fixed θ\thetaθ, and any real qqq, define

J~j(θ,q)=1T∑t=0T−1ρ Πθ(0,mt) Πq(mt,t) rq,mt=max⁡(t−j,0),\widetilde J_j(\theta, q) = \frac{1}{T} \sum_{t=0}^{T-1} \rho \ \Pi_\theta(0, m_t) \ \Pi_q(m_t, t) \ r_q, \qquad m_t = \max(t - j, 0),Jj​(θ,q)=T1​t=0∑T−1​ρ Πθ​(0,mt​) Πq​(mt​,t) rq​,mt​=max(t−j,0),

where ρ\rhoρ is read as a row vector, rqr_qrq​ as a column vector, and each summand is the resulting scalar. In words: in the ttt-th summand the first mtm_tmt​ matrices (indices 0,…,mt−10, \dots, m_t - 10,…,mt​−1) are formed with parameter θ\thetaθ, the remaining matrices (indices mt,…,t−1m_t, \dots, t-1mt​,…,t−1, at most jjj of them) are formed with parameter qqq, and the final vector rqr_qrq​ is formed with parameter qqq. The subtraction t−jt - jt−j is truncated at 000.

Conclusion. All four of the following hold.

(a) An equality of real numbers:

∫ΩGk dμθ=Dk.\int_\Omega G_k \ \mathrm{d}\mu_\theta = D_k .∫Ω​Gk​ dμθ​=Dk​.

(b) The function q↦J~k(θ,q)q \mapsto \widetilde J_k(\theta, q)q↦Jk​(θ,q) from R\mathbb{R}R to R\mathbb{R}R, with the first argument held fixed at θ\thetaθ, is differentiable at the point q=θq = \thetaq=θ (ordinary two-sided derivative), and its derivative there equals DkD_kDk​:

∂∂qJ~k(θ,q)∣q=θ=Dk.\frac{\partial}{\partial q} \widetilde J_k(\theta, q) \Big\vert_{q = \theta} = D_k .∂q∂​Jk​(θ,q)​q=θ​=Dk​.

(c) For every real number qqq with 0≤q0 \le q0≤q and q≤1q \le 1q≤1,

J~T(θ,q)=J(q).\widetilde J_T(\theta, q) = J(q) .JT​(θ,q)=J(q).

Here the truncation index is TTT, not kkk. With j=Tj = Tj=T one has mt=0m_t = 0mt​=0 for every t≤T−1t \le T-1t≤T−1, so Πθ(0,mt)\Pi_\theta(0, m_t)Πθ​(0,mt​) is the identity matrix and the left-hand side is 1T∑t=0T−1ρ Πq(0,t) rq\frac{1}{T} \sum_{t=0}^{T-1} \rho \ \Pi_q(0,t) \ r_qT1​∑t=0T−1​ρ Πq​(0,t) rq​, in which θ\thetaθ no longer appears. Nothing is asserted for qqq outside [0,1][0,1][0,1].

(d) The function J:R→RJ : \mathbb{R} \to \mathbb{R}J:R→R is differentiable at the point q=θq = \thetaq=θ (ordinary two-sided derivative), and its derivative there is

J′(θ)=∫ΩGT dμθ.J'(\theta) = \int_\Omega G_T \ \mathrm{d}\mu_\theta .J′(θ)=∫Ω​GT​ dμθ​.

Here again the window index is TTT, not kkk: GTG_TGT​ uses, for each uuu, all times ttt with u≤t≤T−1u \le t \le T-1u≤t≤T−1.

In (b) and (d), "differentiable at θ\thetaθ with derivative DDD" means f(q)=f(θ)+(q−θ)⋅D+o(q−θ)f(q) = f(\theta) + (q - \theta) \cdot D + o(q - \theta)f(q)=f(θ)+(q−θ)⋅D+o(q−θ) as q→θq \to \thetaq→θ in R\mathbb{R}R. The number kkk occurs only in claims (a) and (b); claims (c) and (d) do not involve kkk. The same real number θ\thetaθ serves as the parameter of the measure μθ\mu_\thetaμθ​, as the parameter in sθs_\thetasθ​, as the first argument of J~\widetilde JJ, and as the point at which the derivatives are taken.

Human review
  • Endorsed by tianyipeng · Oct 12, 2026

  • Endorsed by Wenqian Xing · Oct 12, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me