Any width-m two-layer piecewise-linear network with arbitrary weights that fits n noisy labels below the noise floor has Lip ≳ ε sqrt(n/(m log(m n d/ε))) with high probability on the sphere or Gaussian.
Cambridge University Press, Cambr idge, UK (2015)
2 Pith papers cite this work, alongside 43 external citations. Polarity classification is still indexing.
2
Pith papers citing it
43
external citations · OpenAlex
citation-role summary
background 1
citation-polarity summary
years
2026 2roles
background 1polarities
background 1representative citing papers
Superposition relaxation creates separable estimators for factorable functions that are tighter than McCormick relaxations in numerical tests while providing convergence guarantees.
citing papers explorer
-
A law of robustness for two-layer neural networks with arbitrary weights
Any width-m two-layer piecewise-linear network with arbitrary weights that fits n noisy labels below the noise floor has Lip ≳ ε sqrt(n/(m log(m n d/ε))) with high probability on the sphere or Gaussian.
-
Relaxation via Separable Estimators: Arithmetic and Implementation
Superposition relaxation creates separable estimators for factorable functions that are tighter than McCormick relaxations in numerical tests while providing convergence guarantees.