Shallow Learning Tends to Ridgelet Transform

1 Introduction
2 Setting ▶
3 Basic lemmas of the setting ▶
4 Structure of the minimizers: the null-space effect ▶
5 Sample-size threshold: from the empirical to the population Gibbs minimizer ▶
6 Regularization limits and the geometry of the limit ▶
7 Convergence rates for the regularization limit ▶
8 Dynamics: mean-field Langevin, log-Sobolev inequalities ▶
9 Explicit geometry of the limit
10 Heavy-tailed hidden laws and Sobolev-type representability
11 Material for lean-operator-ridgelet
12 Material for FoML
13 Material for Mathlib
Dependency graph

1 Introduction

This blueprint is generated from the @[blueprint] annotations in the Lean sources. Edit the Lean files, then run ./script/generate.sh.