Skip to main content
PathsFinite-class uniform convergence

Theorem trail

Finite-class uniform convergence

Hoeffding plus a union bound, with every assumption visible.

Record

Research Record

A compact audit view for the theorem before the full source, proof, diagnostic, and failure-mode sections.

Claim scope
theorem
Evidence level
Lean dependency verified
Source precision
reviewed / A
Diagnostics
1 item
Exercises
1 exercise
Failure checks
2 checks

Theorem Statement

If is finite, then for any , with probability at least : Equivalently, a sample of size is -representative with probability at least .

Assumptions

  • iid samples from D
  • bounded loss in [0,1]
  • finite hypothesis class H

Proof Sketch

For each fixed hypothesis, Hoeffding controls the gap between empirical and population risk. Apply the finite union bound over H, then solve the tail probability for epsilon. The whole proof depends on iid sampling, bounded loss, and a finite hypothesis class.

Proof Obligations

Finite class and bounded iid losses

assumption

Verify the hypothesis class is finite, the data are iid, and the loss is bounded before combining per-hypothesis concentration with a union bound.

Infinite classes, adaptive data, or unbounded losses require different capacity or concentration machinery.

Per-hypothesis tail becomes a uniform event

proof step

The proof controls each fixed h with Hoeffding, unions the bad events across H, and solves the combined failure probability for epsilon.

Skipping the union step hides the exact origin of the log |H| sample-complexity penalty.

ERM consequence is not a deep-learning guarantee

boundary

The result is a finite-class uniform-convergence theorem sufficient for ERM analysis, not a blanket explanation of modern overparameterized networks.

Overextending the theorem makes the site look like it is using classical bounds outside their assumptions.

Finite hypothesis class is doing real work

boundary

Do not apply this trail to an infinite or data-dependent class until the finite-class union step has been replaced by a capacity-control argument.

Treating the theorem as a generic generalization guarantee hides the exact point where finiteness is required.

Source Locators

Understanding Machine Learning (2014)

proof

Chapter 4 — Learning via uniform convergence

shalev-shwartz-ben-david-2014:ch4

kind
chapter
review
reviewed / A
extract
manual

Chapter-level support for the finite-class uniform convergence theorem and its Hoeffding plus union-bound proof pattern.

Understanding Machine Learning (2014)

proof

Corollary 4.6 — Finite-class uniform convergence

shalev-shwartz-ben-david-2014:cor4.6

kind
corollary
review
reviewed / A
extract
manual
number
4.6

Anchor: finite hypothesis class

Reviewed corollary-level support for finite-class uniform convergence.

Curated Trace

  1. 1

    Each fixed h gets its own bounded-loss concentration inequality.

    Checkpoint

    Bounded loss in [0,1] is doing real work.

  2. 2
    Union over Hprobability step

    The bad event is a union across hypotheses.

    Checkpoint

    The finite class assumption enters exactly here.

  3. 3

    Solving the combined failure probability gives the sqrt((log |H| + log 1/delta)/n) rate.

    Checkpoint

    Track both delta and |H|; neither is decoration.

  4. 4
    ERM guaranteedownstream

    Uniform convergence controls the excess risk of empirical risk minimization.

    Checkpoint

    The theorem is sufficient for ERM, not a full account of modern deep learning.

Failure Modes

ERM consequence is not a deep-learning guarantee

The result is a finite-class uniform-convergence theorem sufficient for ERM analysis, not a blanket explanation of modern overparameterized networks.

Risk: Overextending the theorem makes the site look like it is using classical bounds outside their assumptions.

Check: The result is a finite-class uniform-convergence theorem sufficient for ERM analysis, not a blanket explanation of modern overparameterized networks.

Finite hypothesis class is doing real work

Do not apply this trail to an infinite or data-dependent class until the finite-class union step has been replaced by a capacity-control argument.

Risk: Treating the theorem as a generic generalization guarantee hides the exact point where finiteness is required.

Check: Do not apply this trail to an infinite or data-dependent class until the finite-class union step has been replaced by a capacity-control argument.