For interpolating kernel machines, minimizing the norm of the ERM solution maximizes stability

DOI10.1142/S0219530522400115OpenAlexW4309259897MaRDI QIDQ5873932

Lorenzo Rosasco, Tomaso Poggio, Akshay Rangamani

Publication date: 10 February 2023

Published in: Analysis and Applications (Search for Journal in Brave)

Full work available at URL: https://doi.org/10.1142/s0219530522400115

zbMATH Keywords

kernel regressionhigh-dimensional statisticsoverparameterizationalgorithmic stabilityminimum-norm interpolation

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05)

Cites Work

• Unnamed Item
• Statistics for high-dimensional data. Methods, theory and applications.
• Interpolation of scattered data: distance matrices and conditionally positive definite functions
• The spectrum of kernel random matrices
• A revisitation of formulae for the Moore-Penrose inverse of modified matrices.
• Surprises in high-dimensional ridgeless least squares interpolation
• Just interpolate: kernel ``ridgeless regression can generalize
• Learning theory: stability is sufficient for generalization and necessary and sufficient for consistency of empirical risk minimization
• Theory of Classification: a Survey of Some Recent Advances
• Support Vector Machines
• 10.1162/153244302760200704
• The Generalization Error of Random Features Regression: Precise Asymptotics and the Double Descent Curve
• Benign overfitting in linear regression
• Reconciling modern machine-learning practice and the classical bias–variance trade-off
• Understanding Machine Learning
• DISTRIBUTION OF EIGENVALUES FOR SOME SETS OF RANDOM MATRICES
• Generalized Inversion of Modified Matrices