Near-Interpolators: Rapid Norm Growth and the Trade-Off between Interpolation and Generalization

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Wang, Yutong, Sonthalia, Rishi, Hu, Wei
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866929273448693760
author Wang, Yutong
Sonthalia, Rishi
Hu, Wei
author_facet Wang, Yutong
Sonthalia, Rishi
Hu, Wei
contents We study the generalization capability of nearly-interpolating linear regressors: $\boldsymbolβ$'s whose training error $τ$ is positive but small, i.e., below the noise floor. Under a random matrix theoretic assumption on the data distribution and an eigendecay assumption on the data covariance matrix $\boldsymbolΣ$, we demonstrate that any near-interpolator exhibits rapid norm growth: for $τ$ fixed, $\boldsymbolβ$ has squared $\ell_2$-norm $\mathbb{E}[\|{\boldsymbolβ}\|_{2}^{2}] = Ω(n^α)$ where $n$ is the number of samples and $α>1$ is the exponent of the eigendecay, i.e., $λ_i(\boldsymbolΣ) \sim i^{-α}$. This implies that existing data-independent norm-based bounds are necessarily loose. On the other hand, in the same regime we precisely characterize the asymptotic trade-off between interpolation and generalization. Our characterization reveals that larger norm scaling exponents $α$ correspond to worse trade-offs between interpolation and generalization. We verify empirically that a similar phenomenon holds for nearly-interpolating shallow neural networks.
format Preprint
id arxiv_https___arxiv_org_abs_2403_07264
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Near-Interpolators: Rapid Norm Growth and the Trade-Off between Interpolation and Generalization
Wang, Yutong
Sonthalia, Rishi
Hu, Wei
Machine Learning
We study the generalization capability of nearly-interpolating linear regressors: $\boldsymbolβ$'s whose training error $τ$ is positive but small, i.e., below the noise floor. Under a random matrix theoretic assumption on the data distribution and an eigendecay assumption on the data covariance matrix $\boldsymbolΣ$, we demonstrate that any near-interpolator exhibits rapid norm growth: for $τ$ fixed, $\boldsymbolβ$ has squared $\ell_2$-norm $\mathbb{E}[\|{\boldsymbolβ}\|_{2}^{2}] = Ω(n^α)$ where $n$ is the number of samples and $α>1$ is the exponent of the eigendecay, i.e., $λ_i(\boldsymbolΣ) \sim i^{-α}$. This implies that existing data-independent norm-based bounds are necessarily loose. On the other hand, in the same regime we precisely characterize the asymptotic trade-off between interpolation and generalization. Our characterization reveals that larger norm scaling exponents $α$ correspond to worse trade-offs between interpolation and generalization. We verify empirically that a similar phenomenon holds for nearly-interpolating shallow neural networks.
title Near-Interpolators: Rapid Norm Growth and the Trade-Off between Interpolation and Generalization
topic Machine Learning
url https://arxiv.org/abs/2403.07264