Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Derivation of Result 6, step 3, p. 357 — averaging the symmetrized step over all base classes gives f_j = ((J−1)/J)(u_j − (1/J)Σ_k u_k)

Proved
AddLogReg.Multiclass.derivation_step_3

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

logitboostmulticlass-classificationnewton-methodp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1

Let J≥2J\ge2J≥2, let ν\nuν be a probability measure on X×{1,…,J}X\times\{1,\dots,J\}X×{1,…,J}, F(x)F(x)F(x) the current fit with model probabilities pj(x)p_j(x)pj​(x), and fix xxx. For a base class bbb let g(b)g^{(b)}g(b) be the diagonal quasi-Newton step of step 2 in base-bbb coordinates (gb(b)=0g^{(b)}_b = 0gb(b)​=0), and convert it to the symmetric parametrization by

fj(b)=gj(b)−1J∑k=1Jgk(b).f^{(b)}_j = g^{(b)}_j - \frac1J\sum_{k=1}^J g^{(b)}_k .fj(b)​=gj(b)​−J1​k=1∑J​gk(b)​.

Write uj=E(yj∗−pj(x)∣x)/(pj(x)(1−pj(x)))u_j = E(y^*_j - p_j(x)\mid x)/\big(p_j(x)(1-p_j(x))\big)uj​=E(yj∗​−pj​(x)∣x)/(pj​(x)(1−pj​(x))). Then the average over all choices of base class is

1J∑b=1Jfj(b)=(J−1J)(E(yj∗−pj(x)∣x)pj(x)(1−pj(x))−1J∑k=1JE(yk∗−pk(x)∣x)pk(x)(1−pk(x))).\frac1J\sum_{b=1}^J f^{(b)}_j = \Big(\frac{J-1}{J}\Big)\bigg(\frac{E(y^*_j - p_j(x)\mid x)}{p_j(x)(1-p_j(x))} - \frac1J\sum_{k=1}^J \frac{E(y^*_k - p_k(x)\mid x)}{p_k(x)(1-p_k(x))}\bigg).J1​b=1∑J​fj(b)​=(JJ−1​)(pj​(x)(1−pj​(x))E(yj∗​−pj​(x)∣x)​−J1​k=1∑J​pk​(x)(1−pk​(x))E(yk∗​−pk​(x)∣x)​).

This is the update of the symmetric J-class LogitBoost: averaging over the base class removes the dependence of the quasi-Newton step on the arbitrary choice of base.

Formalization Note The constants 1/J1/J1/J and (J−1)/J(J-1)/J(J−1)/J are real numbers.

Preamble
import Mathlib
import Definitions.Def_AddLogReg_Multiclass_Setting

open MeasureTheory ProbabilityTheory ENNReal
Formal statement
namespace AddLogReg.Multiclass

theorem derivation_step_3 {J : ℕ} [NeZero J] (hJ : 2 ≤ J) {X : Type*} [MeasurableSpace X]
    (ν : Measure (X × Fin J)) [IsProbabilityMeasure ν] (F : X → Fin J → ℝ) (x : X) (j : Fin J) :
    (1 / (J : ℝ)) * ∑ b, symmetrize (diagStep ν F x b) j =
      (((J : ℝ) - 1) / (J : ℝ)) *
        ((∫ y, (ystar y j - softmax (F x) j) ∂(ν.condKernel x)) /
            (softmax (F x) j * (1 - softmax (F x) j)) -
          (1 / (J : ℝ)) * ∑ k, (∫ y, (ystar y k - softmax (F x) k) ∂(ν.condKernel x)) /
            (softmax (F x) k * (1 - softmax (F x) k))) := by sorry

end AddLogReg.Multiclass
Source
Friedman, Hastie and Tibshirani, Additive logistic regression: a statistical view of boosting, Ann. Statist. 28 (2000), p. 357, Derivation of Result 6, step 3
Human review
  • Endorsed by Shuze Chen · Oct 7, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 7, 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