Stability of the replica symmetric solution in diluted perceptron learning

DOI10.1088/1742-5468/2013/02/L02002zbMATH Open1456.68153arXiv1209.5941WikidataQ57887553 ScholiaQ57887553MaRDI QIDQ3301481FDOQ3301481

Alejandro Lage-Castellanos, Gretel Quintero Angulo, Andrea Pagnani

Publication date: 11 August 2020

Published in: Journal of Statistical Mechanics: Theory and Experiment (Search for Journal in Brave)

Abstract: We study the role played by the dilution in the average behavior of a perceptron model with continuous coupling with the replica method. We analyze the stability of the replica symmetric solution as a function of the dilution field for the generalization and memorization problems. Thanks to a Gardner like stability analysis we show that at any fixed ratio $alpha$ between the number of patterns M and the dimension N of the perceptron ($alpha=M/N$), there exists a critical dilution field $h_c$ above which the replica symmetric ansatz becomes unstable.

Full work available at URL: https://arxiv.org/abs/1209.5941

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05) Neural networks for/in biological studies, artificial life and related topics (92B20) Dynamic critical phenomena in statistical mechanics (82C27)

Cites Work

• 10.1162/153244303322753616
• Title not available (Why is that?)
• Compressed sensing
• Title not available (Why is that?)
• Feature extraction. Foundations and applications. Papers from NIPS 2003 workshop on feature extraction, Whistler, BC, Canada, December 11--13, 2003. With CD-ROM.
• Statistical mechanics of learning
• The space of interactions in neural network models
• Quenched versus annealed dilution in neural networks

Cited In (2)

• Optimal capacity of graded-response perceptrons: a replica-symmetry-breaking solution
• Maximal stability in unsupervised learning

This page was built for publication: Stability of the replica symmetric solution in diluted perceptron learning

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3301481)