Free dynamics of feature learning processes

DOI10.1007/s10955-022-03064-5OpenAlexW4310762320MaRDI QIDQ2679634

Cyril Furtlehner

Publication date: 23 January 2023

Published in: Journal of Statistical Physics (Search for Journal in Brave)

Full work available at URL: https://arxiv.org/abs/2210.10702

zbMATH Keywords

dynamical systemsneural networksrandom matrix theorymachine learning

Mathematics Subject Classification ID

Linear inference, regression (62Jxx) Artificial intelligence (68Txx) Probability theory on algebraic and topological structures (60Bxx)

Cites Work

• Unnamed Item
• Unnamed Item
• Unnamed Item
• Free probability and random matrices
• Cleaning large correlation matrices: tools from random matrix theory
• Eigenvectors of some large sample covariance matrix ensembles
• Limit laws for random matrices and free products
• A random matrix approach to neural networks
• Why does deep and cheap learning work so well?
• High-dimensional asymptotics of prediction: ridge regression and classification
• Thermodynamics of restricted Boltzmann machines and related learning dynamics
• High-dimensional dynamics of generalization error in neural networks
• Surprises in high-dimensional ridgeless least squares interpolation
• Deterministic equivalents for certain functionals of large random matrices
• Field Theories for Learning Probability Distributions
• 10.1162/153244302760200704
• Generalisation error in learning with random features and the hidden manifold model*
• Learning curves of generic features maps for realistic datasets with a teacher-student model*
• The Generalization Error of Random Features Regression: Precise Asymptotics and the Double Descent Curve
• The unreasonable effectiveness of deep learning in artificial intelligence
• Benign overfitting in linear regression
• Random Matrix Methods for Machine Learning
• Reconciling modern machine-learning practice and the classical bias–variance trade-off
• DISTRIBUTION OF EIGENVALUES FOR SOME SETS OF RANDOM MATRICES
• Scaling description of generalization with number of parameters in deep learning
• Asymptotic learning curves of kernel methods: empirical data versus teacher–student paradigm
• Wide neural networks of any depth evolve as linear models under gradient descent ^*
• Dynamics of stochastic gradient descent for two-layer neural networks in the teacher–student setup*
• Gaussian-spherical restricted Boltzmann machines
• Deep learning: a statistical viewpoint

This page was built for publication: Free dynamics of feature learning processes