Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Stationary Gaussian model and standard DKF recursion

Definition
DKFAsymptotics_Filter

by burkh4rt · Sep 15, 2026 · Mathlib 0df444a (Lean v4.33.1)

gaussiankalman-filterprobabilitystatistics

The model consists of A,Γ,SA,\Gamma,SA,Γ,S with positive-definite noise and stationary covariances and S=ASA⊤+ΓS=ASA^\top+\GammaS=ASA⊤+Γ. It defines the stationary density, Gaussian transition, prediction integral, and discriminative update uτs/pu\tau s/puτs/p normalized by its L1L^1L1 norm. Update validity means a measurable weight with finite, strictly positive normalizer. The standard DKF matrix step uses prediction covariance AVA⊤+ΓAVA^\top+\GammaAVA⊤+Γ, updated precision U−1+(AVA⊤+Γ)−1−S−1U^{-1}+(AVA^\top+\Gamma)^{-1}-S^{-1}U−1+(AVA⊤+Γ)−1−S−1, its inverse covariance, and the corresponding precision-weighted mean. The recursive DKF starts at mean zero and covariance SSS and incorporates the observation functions f,Qf,Qf,Q at each positive time. Matrix expressions are total; a separate validity predicate requires invertible precision and positive-definite updated covariance. No universal step-validity condition is imposed by the recursion.

Definition code
import Definitions.Def_DKFAsymptotics_Densities

noncomputable section
open MeasureTheory Filter Matrix
open scoped Topology ENNReal

namespace DKFAsymptotics

structure Model (d : ℕ) where
  A : Covariance d
  Γ : Covariance d
  S : Covariance d
  Γ_posDef : Γ.PosDef
  S_posDef : S.PosDef
  stationary : S = A * S * A.transpose + Γ

variable {d : ℕ}

def prior (M : Model d) : Density d := gaussianPDF 0 M.S

def transition (M : Model d) (y z : State d) : ℝ :=
  gaussianPDF (M.A *ᵥ y) M.Γ z

def predictDensity (M : Model d) (s : Density d) : Density d :=
  fun z => ∫ y, transition M y z * s y ∂volume

def updateWeight (M : Model d) (s u : Density d) : Density d :=
  fun z => u z * predictDensity M s z / prior M z

def UpdateDefined (M : Model d) (s u : Density d) : Prop :=
  AEStronglyMeasurable (updateWeight M s u) volume ∧
    0 < l1Norm (updateWeight M s u) ∧ l1Norm (updateWeight M s u) < ⊤

def updateDensity (M : Model d) (s u : Density d) : Density d :=
  fun z => updateWeight M s u z / (l1Norm (updateWeight M s u)).toReal

structure GaussianState (d : ℕ) where
  mean : State d
  covariance : Covariance d

def predictedCovariance (M : Model d) (V : Covariance d) : Covariance d :=
  M.A * V * M.A.transpose + M.Γ

def updatePrecision (M : Model d) (V U : Covariance d) : Covariance d :=
  U⁻¹ + (predictedCovariance M V)⁻¹ - M.S⁻¹

def dkfStep (M : Model d) (previous observation : GaussianState d) : GaussianState d :=
  let G := predictedCovariance M previous.covariance
  let T := (updatePrecision M previous.covariance observation.covariance)⁻¹
  ⟨T *ᵥ (observation.covariance⁻¹ *ᵥ observation.mean +
    G⁻¹ *ᵥ (M.A *ᵥ previous.mean)), T⟩

def ValidStep (M : Model d) (previous observation : GaussianState d) : Prop :=
  IsUnit (updatePrecision M previous.covariance observation.covariance).det ∧
    (dkfStep M previous observation).covariance.PosDef

def dkf {X : Type} (M : Model d) (f : X → State d) (Q : X → Covariance d)
    (observations : ℕ+ → X) : ℕ → GaussianState d
  | 0 => ⟨0, M.S⟩
  | t + 1 => dkfStep M (dkf M f Q observations t)
      ⟨f (observations ⟨t + 1, Nat.succ_pos t⟩), Q (observations ⟨t + 1, Nat.succ_pos t⟩)⟩

end DKFAsymptotics
Source
M. C. Burkhart, D. M. Brandman, B. Franco, L. R. Hochberg and M. T. Harrison, The Discriminative Kalman Filter for Bayesian Filtering with Nonlinear and Nongaussian Observation Models, Neural Computation 32(5), 969–1017 (2020), doi:10.1162/neco_a_01275; https://burkh4rt.github.io/pubs/Burkhart-et-al-2020-Neural-Comput.pdf; Section 2.1, equations (2.1), (2.4)–(2.7), pp. 972–974; Appendix B, equation (B.1), p. 1003; Lemma 1, pp. 1007–1008.
Read-back

What the Lean code literally says, in plain math · GPT-6 (Codex independent sub-agent)

Independent read-back: filter declarations

Scope: the thirteen source declarations in Definitions/Def_DKFAsymptotics_Filter.lean. Only that file, its imported density definitions, and relevant Mathlib definitions were used. In the paragraphs below, vectors have coordinates indexed by the finite set with ddd elements, matrices are real d×dd\times dd×d matrices, and dzdzdz is product Lebesgue measure on Rd\mathbb R^dRd. A “positive definite” real matrix is symmetric and has strictly positive quadratic form on every nonzero vector. Dimension d=0d=0d=0 is allowed; its vector space is a singleton of Lebesgue mass one and its unique empty matrix has determinant one. Matrix inversion means the ordinary inverse for an invertible matrix and the zero matrix otherwise. Real division by zero is zero. These conventions are also stated where they affect a declaration's domain or result.

Model

For each natural number d≥0d\geq0d≥0, a model consists of three real d×dd\times dd×d matrices A,Γ,SA,\Gamma,SA,Γ,S, proofs that Γ\GammaΓ and SSS are symmetric and satisfy xTΓx>0x^\mathsf T\Gamma x>0xTΓx>0 and xTSx>0x^\mathsf TSx>0xTSx>0 for every nonzero x∈Rdx\in\mathbb R^dx∈Rd, and a proof of the exact matrix equality S=ASAT+ΓS=ASA^\mathsf T+\GammaS=ASAT+Γ. There are no separately imposed symmetry, invertibility, or positive definiteness conditions on AAA, and the matrix type itself imposes none on any of the three matrices; the two positivity conditions are the explicit additional fields. No random variables, probability measures, observation space, or observation law are fields of this structure. For d=0d=0d=0, all three matrices are the unique empty matrix, symmetry and the displayed equality hold, and the strict quadratic-form conditions are vacuous because there are no nonzero vectors.

prior

For every natural number d≥0d\geq0d≥0 and every model consisting of real d×dd\times dd×d matrices A,Γ,SA,\Gamma,SA,Γ,S with Γ,S\Gamma,SΓ,S symmetric positive definite and S=ASAT+ΓS=ASA^\mathsf T+\GammaS=ASAT+Γ, the prior is the real-valued function on Rd\mathbb R^dRd given by pM(z)=((2π)ddet⁡S)−1exp⁡(−zTS−1z/2)p_M(z)=(\sqrt{(2\pi)^d\det S})^{-1}\exp(-z^\mathsf TS^{-1}z/2)pM​(z)=((2π)ddetS​)−1exp(−zTS−1z/2). This is a definition of a function, with no additional probability-density predicate bundled into its return value. Matrix inversion is total, being zero on singular matrices, and the real square root is zero on nonpositive arguments, but the model's positive definiteness of SSS gives positive determinant and therefore prevents a zero denominator in this formula. If d=0d=0d=0, the unique input vector has quadratic form zero, the determinant is one, and the function value is one.

transition

For every natural number d≥0d\geq0d≥0, every model M=(A,Γ,S)M=(A,\Gamma,S)M=(A,Γ,S) of real matrices with symmetric positive definite Γ,S\Gamma,SΓ,S and S=ASAT+ΓS=ASA^\mathsf T+\GammaS=ASAT+Γ, and every pair of vectors y,z∈Rdy,z\in\mathbb R^dy,z∈Rd, the transition value is the real number kM(y,z)=((2π)ddet⁡Γ)−1exp⁡(−(z−Ay)TΓ−1(z−Ay)/2)k_M(y,z)=(\sqrt{(2\pi)^d\det\Gamma})^{-1}\exp(-(z-Ay)^\mathsf T\Gamma^{-1}(z-Ay)/2)kM​(y,z)=((2π)ddetΓ​)−1exp(−(z−Ay)TΓ−1(z−Ay)/2). The arguments y,zy,zy,z are arbitrary vectors, and this declaration returns a real number rather than a probability measure or a proof about an integral. Positive definiteness of Γ\GammaΓ gives a positive determinant and a nonzero normalizing denominator, despite the underlying total conventions that matrix inverses of singular matrices and real reciprocals of zero are zero. For d=0d=0d=0, there is one pair of vectors and its transition value is one.

predictDensity

For every natural number d≥0d\geq0d≥0, every model M=(A,Γ,S)M=(A,\Gamma,S)M=(A,Γ,S) with symmetric positive definite real matrices Γ,S\Gamma,SΓ,S satisfying S=ASAT+ΓS=ASA^\mathsf T+\GammaS=ASAT+Γ, and every real-valued function s:Rd→Rs:\mathbb R^d\to\mathbb Rs:Rd→R, the predicted function is qM,s(z)=∫Rd((2π)ddet⁡Γ)−1exp⁡(−(z−Ay)TΓ−1(z−Ay)/2)s(y) dyq_{M,s}(z)=\int_{\mathbb R^d}(\sqrt{(2\pi)^d\det\Gamma})^{-1}\exp(-(z-Ay)^\mathsf T\Gamma^{-1}(z-Ay)/2)s(y)\,dyqM,s​(z)=∫Rd​((2π)ddetΓ​)−1exp(−(z−Ay)TΓ−1(z−Ay)/2)s(y)dy. The integral is the real-valued Bochner integral with respect to Lebesgue measure: at any fixed zzz where the displayed integrand is not integrable, its value is defined to be zero. There is no hypothesis that sss is measurable, integrable, nonnegative, normalized, or Gaussian, and no such property is included in the return type. Thus the definition accepts signed and nonmeasurable functions as well as probability densities. When d=0d=0d=0, the space contains a single point eee of mass one, the kernel equals one, and qM,s(e)=s(e)q_{M,s}(e)=s(e)qM,s​(e)=s(e).

updateWeight

For every natural number d≥0d\geq0d≥0, every model M=(A,Γ,S)M=(A,\Gamma,S)M=(A,Γ,S) with symmetric positive definite real matrices Γ,S\Gamma,SΓ,S satisfying S=ASAT+ΓS=ASA^\mathsf T+\GammaS=ASAT+Γ, and arbitrary real-valued functions s,u:Rd→Rs,u:\mathbb R^d\to\mathbb Rs,u:Rd→R, the weight function is wM,s,u(z)=u(z)qM,s(z)/pM(z)w_{M,s,u}(z)=u(z)q_{M,s}(z)/p_M(z)wM,s,u​(z)=u(z)qM,s​(z)/pM​(z), where pM(z)=((2π)ddet⁡S)−1exp⁡(−zTS−1z/2)p_M(z)=(\sqrt{(2\pi)^d\det S})^{-1}\exp(-z^\mathsf TS^{-1}z/2)pM​(z)=((2π)ddetS​)−1exp(−zTS−1z/2) and qM,s(z)=∫Rd((2π)ddet⁡Γ)−1exp⁡(−(z−Ay)TΓ−1(z−Ay)/2)s(y) dyq_{M,s}(z)=\int_{\mathbb R^d}(\sqrt{(2\pi)^d\det\Gamma})^{-1}\exp(-(z-Ay)^\mathsf T\Gamma^{-1}(z-Ay)/2)s(y)\,dyqM,s​(z)=∫Rd​((2π)ddetΓ​)−1exp(−(z−Ay)TΓ−1(z−Ay)/2)s(y)dy. The latter is a Bochner integral defined to be zero when its integrand is not integrable. No measurability, sign, normalization, or Gaussian assumption is made on either input function, and neither an update-defined condition nor any other condition is required to form this weight. Real division is total and gives zero for a zero denominator, although pM(z)>0p_M(z)>0pM​(z)>0 for the model matrices here. In dimension zero, at the unique point eee, the formula is wM,s,u(e)=u(e)s(e)w_{M,s,u}(e)=u(e)s(e)wM,s,u​(e)=u(e)s(e).

UpdateDefined

For every natural number d≥0d\geq0d≥0, every model M=(A,Γ,S)M=(A,\Gamma,S)M=(A,Γ,S) with symmetric positive definite real matrices Γ,S\Gamma,SΓ,S satisfying S=ASAT+ΓS=ASA^\mathsf T+\GammaS=ASAT+Γ, and arbitrary real-valued functions s,u:Rd→Rs,u:\mathbb R^d\to\mathbb Rs,u:Rd→R, this proposition holds exactly when the real function w(z)=u(z)q(z)/p(z)w(z)=u(z)q(z)/p(z)w(z)=u(z)q(z)/p(z) is equal Lebesgue-almost everywhere to a strongly measurable function and its extended nonnegative integral satisfies 0<∫Rd−∣w(z)∣ dz<∞0<\int^-_{\mathbb R^d}|w(z)|\,dz<\infty0<∫Rd−​∣w(z)∣dz<∞; here p(z)=((2π)ddet⁡S)−1exp⁡(−zTS−1z/2)p(z)=(\sqrt{(2\pi)^d\det S})^{-1}\exp(-z^\mathsf TS^{-1}z/2)p(z)=((2π)ddetS​)−1exp(−zTS−1z/2), and q(z)=∫Rd((2π)ddet⁡Γ)−1exp⁡(−(z−Ay)TΓ−1(z−Ay)/2)s(y) dyq(z)=\int_{\mathbb R^d}(\sqrt{(2\pi)^d\det\Gamma})^{-1}\exp(-(z-Ay)^\mathsf T\Gamma^{-1}(z-Ay)/2)s(y)\,dyq(z)=∫Rd​((2π)ddetΓ​)−1exp(−(z−Ay)TΓ−1(z−Ay)/2)s(y)dy, with the inner real Bochner integral set to zero if its integrand is not integrable. The notation for the outer integral embeds the nonnegative real number ∣w(z)∣|w(z)|∣w(z)∣ into the extended nonnegative reals before integration. Strong measurability here means representability as a pointwise limit of measurable simple functions; only an almost-everywhere equality with such a representative is required of www. The proposition does not require nonnegativity of w,s,uw,s,uw,s,u, integral one, Gaussian form, or integrability of each inner prediction integrand; its norm condition concerns the absolute value of the resulting weight. Zero weights fail the strict lower bound, and an infinite absolute-value integral fails the strict upper bound. For d=0d=0d=0, every function on the singleton is measurable and finite-valued, so the proposition is exactly u(e)s(e)≠0u(e)s(e)\ne0u(e)s(e)=0.

updateDensity

For every natural number d≥0d\geq0d≥0, every model M=(A,Γ,S)M=(A,\Gamma,S)M=(A,Γ,S) with symmetric positive definite real matrices Γ,S\Gamma,SΓ,S satisfying S=ASAT+ΓS=ASA^\mathsf T+\GammaS=ASAT+Γ, and arbitrary real-valued functions s,u:Rd→Rs,u:\mathbb R^d\to\mathbb Rs,u:Rd→R, the updated function is r(z)=w(z)/toReal⁡(N)r(z)=w(z)/\operatorname{toReal}(N)r(z)=w(z)/toReal(N), where w(z)=u(z)q(z)/p(z)w(z)=u(z)q(z)/p(z)w(z)=u(z)q(z)/p(z), N=∫Rd−∣w(z)∣ dz∈[0,∞]N=\int^-_{\mathbb R^d}|w(z)|\,dz\in[0,\infty]N=∫Rd−​∣w(z)∣dz∈[0,∞], p(z)=((2π)ddet⁡S)−1exp⁡(−zTS−1z/2)p(z)=(\sqrt{(2\pi)^d\det S})^{-1}\exp(-z^\mathsf TS^{-1}z/2)p(z)=((2π)ddetS​)−1exp(−zTS−1z/2), and q(z)=∫Rd((2π)ddet⁡Γ)−1exp⁡(−(z−Ay)TΓ−1(z−Ay)/2)s(y) dyq(z)=\int_{\mathbb R^d}(\sqrt{(2\pi)^d\det\Gamma})^{-1}\exp(-(z-Ay)^\mathsf T\Gamma^{-1}(z-Ay)/2)s(y)\,dyq(z)=∫Rd​((2π)ddetΓ​)−1exp(−(z−Ay)TΓ−1(z−Ay)/2)s(y)dy, with any nonintegrable inner Bochner integral defined to be zero. The extended integral defining NNN uses the nonnegative real absolute value embedded into the extended nonnegative reals. The conversion toReal⁡\operatorname{toReal}toReal returns the corresponding real for a finite NNN and returns zero when N=∞N=\inftyN=∞; real division by zero is zero. Consequently, N=0N=0N=0 and N=∞N=\inftyN=∞ both make rrr identically zero, including at any exceptional points where w≠0w\ne0w=0. No update-defined, measurability, integrability, sign, or normalization assumption is a parameter of this definition, and absolute-value normalization does not remove negative values of www. For d=0d=0d=0, the value at the unique point eee is u(e)s(e)/∣u(e)s(e)∣u(e)s(e)/|u(e)s(e)|u(e)s(e)/∣u(e)s(e)∣, with value zero when the product is zero.

GaussianState

For every natural number d≥0d\geq0d≥0, a Gaussian state is exactly a pair consisting of a vector m∈Rdm\in\mathbb R^dm∈Rd, called its mean, and an arbitrary real d×dd\times dd×d matrix VVV, called its covariance. The structure imposes no symmetry, positivity, positive definiteness, invertibility, or probability-distribution condition on the matrix, and contains no density, measure, random variable, or assertion linking the two fields to a probability law. In particular, singular, nonsymmetric, and indefinite matrices are allowed whenever their dimensions permit them. For d=0d=0d=0, both fields are unique, so there is exactly one such pair.

predictedCovariance

For every natural number d≥0d\geq0d≥0, every model M=(A,Γ,S)M=(A,\Gamma,S)M=(A,Γ,S) of real d×dd\times dd×d matrices with symmetric positive definite Γ,S\Gamma,SΓ,S and S=ASAT+ΓS=ASA^\mathsf T+\GammaS=ASAT+Γ, and every real d×dd\times dd×d matrix VVV, this definition returns GM(V)=AVAT+ΓG_M(V)=AVA^\mathsf T+\GammaGM​(V)=AVAT+Γ. There is no positivity, symmetry, or invertibility hypothesis on VVV, and the return value is an unrestricted matrix rather than a matrix bundled with positive definiteness. The formula uses matrix multiplication in the indicated order and the transpose of AAA. For d=0d=0d=0, it returns the unique empty matrix.

updatePrecision

For every natural number d≥0d\geq0d≥0, every model M=(A,Γ,S)M=(A,\Gamma,S)M=(A,Γ,S) with symmetric positive definite real matrices Γ,S\Gamma,SΓ,S satisfying S=ASAT+ΓS=ASA^\mathsf T+\GammaS=ASAT+Γ, and arbitrary real d×dd\times dd×d matrices V,UV,UV,U, the precision expression is the matrix PM(V,U)=U−1+(AVAT+Γ)−1−S−1P_M(V,U)=U^{-1}+(AVA^\mathsf T+\Gamma)^{-1}-S^{-1}PM​(V,U)=U−1+(AVAT+Γ)−1−S−1. Each matrix inverse is the usual inverse if its determinant is nonzero and the zero matrix otherwise. The model assumption ensures that SSS is invertible, but no positivity, symmetry, or invertibility hypothesis is made on VVV, on UUU, on AVAT+ΓAVA^\mathsf T+\GammaAVAT+Γ, or on the resulting precision expression. Thus a singular UUU or a singular predicted matrix contributes a zero inverse term, and the definition still returns a matrix. For d=0d=0d=0, it returns the unique empty matrix, whose determinant is one.

dkfStep

For every natural number d≥0d\geq0d≥0, every model M=(A,Γ,S)M=(A,\Gamma,S)M=(A,Γ,S) with symmetric positive definite real matrices Γ,S\Gamma,SΓ,S satisfying S=ASAT+ΓS=ASA^\mathsf T+\GammaS=ASAT+Γ, and every two pairs (m,V)(m,V)(m,V) and (b,U)(b,U)(b,U) of a vector in Rd\mathbb R^dRd and an arbitrary real d×dd\times dd×d matrix, called the previous and observation states respectively, define G=AVAT+ΓG=AVA^\mathsf T+\GammaG=AVAT+Γ, P=U−1+G−1−S−1P=U^{-1}+G^{-1}-S^{-1}P=U−1+G−1−S−1, and T=P−1T=P^{-1}T=P−1; the step returns the pair (T(U−1b+G−1Am),T)(T(U^{-1}b+G^{-1}Am),T)(T(U−1b+G−1Am),T). All matrix inverses are ordinary inverses at invertible matrices and zero at singular matrices. No positive definiteness, symmetry, invertibility, or validity condition is imposed on either input state, on GGG, on PPP, or on the output; this is a total definition on all the stated inputs. If PPP is singular, then T=0T=0T=0 and the returned vector and matrix are both zero. The term involving S−1S^{-1}S−1 appears only in PPP; the formula contains no additional vector term using S−1S^{-1}S−1. For d=0d=0d=0, the result is the unique pair of an empty vector and an empty matrix.

ValidStep

For every natural number d≥0d\geq0d≥0, every model M=(A,Γ,S)M=(A,\Gamma,S)M=(A,Γ,S) with symmetric positive definite real matrices Γ,S\Gamma,SΓ,S satisfying S=ASAT+ΓS=ASA^\mathsf T+\GammaS=ASAT+Γ, and arbitrary previous and observation pairs (m,V)(m,V)(m,V), (b,U)(b,U)(b,U) of vectors and real matrices, this proposition is the conjunction that det⁡P\det PdetP is a unit in R\mathbb RR, equivalently det⁡P≠0\det P\ne0detP=0, and that P−1P^{-1}P−1 is symmetric positive definite, where P=U−1+(AVAT+Γ)−1−S−1P=U^{-1}+(AVA^\mathsf T+\Gamma)^{-1}-S^{-1}P=U−1+(AVAT+Γ)−1−S−1. The latter matrix is the covariance field of the pair (P−1(U−1b+(AVAT+Γ)−1Am),P−1)(P^{-1}(U^{-1}b+(AVA^\mathsf T+\Gamma)^{-1}Am),P^{-1})(P−1(U−1b+(AVAT+Γ)−1Am),P−1) returned by the step. Matrix inverses use zero at singular inputs, so UUU and AVAT+ΓAVA^\mathsf T+\GammaAVAT+Γ remain allowed to be singular; the proposition imposes no separate positive definiteness or invertibility condition on them or on VVV. It imposes no condition on the two mean vectors, and neither a probability density nor a normalization integral occurs in it. In dimension zero, det⁡P=1\det P=1detP=1, symmetry holds, and strict positivity on nonzero vectors is vacuous, so the proposition holds for the unique state pairs.

dkf

For every natural number d≥0d\geq0d≥0, every type XXX, every model M=(A,Γ,S)M=(A,\Gamma,S)M=(A,Γ,S) with symmetric positive definite real matrices Γ,S\Gamma,SΓ,S satisfying S=ASAT+ΓS=ASA^\mathsf T+\GammaS=ASAT+Γ, every two functions f:X→Rdf:X\to\mathbb R^df:X→Rd and Q:X→Rd×dQ:X\to\mathbb R^{d\times d}Q:X→Rd×d, and every sequence x:N>0→Xx:\mathbb N_{>0}\to Xx:N>0​→X indexed by strictly positive natural numbers, this definition returns a sequence of vector-matrix pairs (mt,Vt)(m_t,V_t)(mt​,Vt​) for all natural times t≥0t\geq0t≥0. Its initial value is (m0,V0)=(0,S)(m_0,V_0)=(0,S)(m0​,V0​)=(0,S). For every natural t≥0t\geq0t≥0, put Gt=AVtAT+ΓG_t=AV_tA^\mathsf T+\GammaGt​=AVt​AT+Γ, Ut+1=Q(xt+1)U_{t+1}=Q(x_{t+1})Ut+1​=Q(xt+1​), and Tt+1=(Ut+1−1+Gt−1−S−1)−1T_{t+1}=(U_{t+1}^{-1}+G_t^{-1}-S^{-1})^{-1}Tt+1​=(Ut+1−1​+Gt−1​−S−1)−1; then Vt+1=Tt+1V_{t+1}=T_{t+1}Vt+1​=Tt+1​ and mt+1=Tt+1(Ut+1−1f(xt+1)+Gt−1Amt)m_{t+1}=T_{t+1}(U_{t+1}^{-1}f(x_{t+1})+G_t^{-1}Am_t)mt+1​=Tt+1​(Ut+1−1​f(xt+1​)+Gt−1​Amt​). The observation at index t+1t+1t+1 exists because that index is positive; there is no observation at index zero used in the initialization. All matrix inverses are total and become zero for singular matrices, and no positivity or symmetry of Q(x)Q(x)Q(x), valid-step condition, measurability, probabilistic assumption, or regularity assumption on X,f,Q,xX,f,Q,xX,f,Q,x is required. In particular, the recursion still returns values if a precision expression is singular, in which case that step's vector and covariance are zero. Although XXX itself is unrestricted, supplying xxx entails that XXX is inhabited, since x1∈Xx_1\in Xx1​∈X; no complete input tuple exists when XXX is empty. In dimension zero, the sequence is constantly the unique pair of empty vector and empty matrix.

Human review
  • Endorsed by Shuze Chen · Sep 16, 2026

  • Endorsed by burkh4rt · Sep 16, 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