Machine learning meets physics: A two-way street

H Levine, Y Tu - Proceedings of the National Academy of Sciences, 2024 - pnas.org
This article introduces a special issue on the interaction between the rapidly expanding field
of machine learning and ongoing research in physics. The first half of the papers in this …

Statistical mechanics of transfer learning in fully-connected networks in the proportional limit

A Ingrosso, R Pacelli, P Rotondo, F Gerace - arXiv preprint arXiv …, 2024 - arxiv.org
Transfer learning (TL) is a well-established machine learning technique to boost the
generalization performance on a specific (target) task using information gained from a …

Deep learning-driven predictive tools for damage prediction and optimization in composite hydrogen storage tanks

M Nachtane, MEF Idrissi, M Tarfaoui, Y Qarssis… - Composites …, 2024 - Elsevier
This research presents a comprehensive framework for predicting the damage in lightweight
composite high-pressure hydrogen storage tanks and optimizes their design to prevent …

[HTML][HTML] Practice Reshapes the Geometry and Dynamics of Task-tailored Representations

A Kikumoto, K Shibata, T Nishio, D Badre - bioRxiv, 2024 - pmc.ncbi.nlm.nih.gov
Extensive practice makes task performance more efficient and precise, leading to
automaticity. However, theories of automaticity differ on which levels of task representations …

Transitional probabilities outweigh frequency of occurrence in statistical learning of simultaneously presented visual shapes

AD Endress - Memory & Cognition, 2024 - Springer
Statistical learning is a mechanism for detecting associations among co-occurring elements
in many domains and species. A key controversy is whether it leads to memory for discrete …

Simplified derivations for high-dimensional convex learning problems

DG Clark, H Sompolinsky - arXiv preprint arXiv:2412.01110, 2024 - arxiv.org
Statistical physics provides tools for analyzing high-dimensional problems in machine
learning and theoretical neuroscience. These calculations, particularly those using the …

Kernel Shape Renormalization In Bayesian Shallow Networks: a Gaussian Process Perspective

R Pacelli, L Giambagli… - 2024 IEEE Workshop on …, 2024 - ieeexplore.ieee.org
The Bayesian approach has proven to be a valuable tool for analytical inspection of neural
networks. Recent theoretical advances have led to the development of an effective statistical …

Delays in generalization match delayed changes in representational geometry

X Zheng, K Daruwalla, AS Benjamin, D Klindt - UniReps: 2nd Edition of the … - openreview.net
Delayed generalization, also known as``grokking'', has emerged as a well-replicated
phenomenon in overparameterized neural networks. Recent theoretical works associated …