Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Li, Jiping, Sonthalia, Rishi
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2510.01414
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866914071384686592
author Li, Jiping
Sonthalia, Rishi
author_facet Li, Jiping
Sonthalia, Rishi
contents This paper analyzes the generalization error of minimum-norm interpolating solutions in linear regression using spiked covariance data models. The paper characterizes how varying spike strengths and target-spike alignments can affect risk, especially in overparameterized settings. The study presents an exact expression for the generalization error, leading to a comprehensive classification of benign, tempered, and catastrophic overfitting regimes based on spike strength, the aspect ratio $c=d/n$ (particularly as $c \to \infty$), and target alignment. Notably, in well-specified aligned problems, increasing spike strength can surprisingly induce catastrophic overfitting before achieving benign overfitting. The paper also reveals that target-spike alignment is not always advantageous, identifying specific, sometimes counterintuitive, conditions for its benefit or detriment. Alignment with the spike being detrimental is empirically demonstrated to persist in nonlinear models.
format Preprint
id arxiv_https___arxiv_org_abs_2510_01414
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Risk Phase Transitions in Spiked Regression: Alignment Driven Benign and Catastrophic Overfitting
Li, Jiping
Sonthalia, Rishi
Machine Learning
Artificial Intelligence
This paper analyzes the generalization error of minimum-norm interpolating solutions in linear regression using spiked covariance data models. The paper characterizes how varying spike strengths and target-spike alignments can affect risk, especially in overparameterized settings. The study presents an exact expression for the generalization error, leading to a comprehensive classification of benign, tempered, and catastrophic overfitting regimes based on spike strength, the aspect ratio $c=d/n$ (particularly as $c \to \infty$), and target alignment. Notably, in well-specified aligned problems, increasing spike strength can surprisingly induce catastrophic overfitting before achieving benign overfitting. The paper also reveals that target-spike alignment is not always advantageous, identifying specific, sometimes counterintuitive, conditions for its benefit or detriment. Alignment with the spike being detrimental is empirically demonstrated to persist in nonlinear models.
title Risk Phase Transitions in Spiked Regression: Alignment Driven Benign and Catastrophic Overfitting
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2510.01414