Prove2Me
Navigate
DiscoverCollectionsFormalpediaBlogsUsersMomentumMy Missions+
Prove2Me
⌕
Log in
← Formalpedia

Corollary 4.4: uniform convergence with m^UC_H implies agnostic PAC learnability by ERM with m_H(ε, δ) ≤ m^UC_H(ε/2, δ)

Proved
UnderstandingML.uc_implies_agnostic_pac

by naimengye · Sep 24, 2026 · Mathlib 0df444a (Lean v4.33.1)

agnostic-pacermuniform-convergence

Corollary 4.4. If a class HHH has the uniform convergence property with a function mHUCm^{UC}_HmHUC​ then the class is agnostically PAC learnable with the sample complexity mH(ϵ,δ)≤mHUC(ϵ/2,δ)m_H(\epsilon, \delta) \le m^{UC}_H(\epsilon/2, \delta)mH​(ϵ,δ)≤mHUC​(ϵ/2,δ). Furthermore, in that case, the ERMHERM_HERMH​ paradigm is a successful agnostic PAC learner for HHH.

Formally: if HHH has the uniform convergence property with mHUCm^{UC}_HmHUC​, then every ERM learner for HHH satisfies the agnostic PAC guarantee (Definition 3.4) with the sample-complexity function (ϵ,δ)↦mHUC(ϵ/2,δ)(\epsilon, \delta) \mapsto m^{UC}_H(\epsilon/2, \delta)(ϵ,δ)↦mHUC​(ϵ/2,δ), and HHH is agnostic PAC learnable provided an ERM learner exists.

Preamble
import Definitions.Def_UnderstandingML_Framework

open MeasureTheory
Formal statement
namespace UnderstandingML

/-- **Corollary 4.4** (p. 55). If a class `H` has the uniform convergence property with a function
`m^{UC}_H`, then the class is agnostically PAC learnable with the sample complexity
`m_H(ε, δ) ≤ m^{UC}_H(ε/2, δ)`; furthermore, in that case the `ERM_H` paradigm is a successful
agnostic PAC learner for `H`: every ERM learner satisfies the agnostic PAC guarantee with the
function `(ε, δ) ↦ m^{UC}_H(ε/2, δ)`, and `H` is agnostic PAC learnable as soon as an ERM learner
exists. -/
theorem uc_implies_agnostic_pac {Z : Type*} [MeasurableSpace Z] {Hyp : Type*}
    (loss : Hyp → Z → ℝ) (H : Set Hyp) (mUC : ℝ → ℝ → ℕ)
    (hUC : HasUniformConvergenceWith loss H mUC) :
    (∀ A : Learner Z Hyp, IsERMLearner loss H A →
      IsAgnosticPACWith loss H A (fun ε δ ↦ mUC (ε / 2) δ)) ∧
    ((∃ A : Learner Z Hyp, IsERMLearner loss H A) → AgnosticPACLearnable loss H) := by sorry

end UnderstandingML
Source
Shalev-Shwartz and Ben-David, Understanding Machine Learning: From Theory to Algorithms, Cambridge University Press 2014, doi:10.1017/CBO9781107298019, §4.1 p. 55, Corollary 4.4 (from Lemma 4.2 and Definition 4.3)
Human review
  • Endorsed by Shuze Chen · Sep 25, 2026

    Confirmed by the moderator at approval.

  • Endorsed by naimengye · Sep 25, 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 paper
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me