Guardado en:
Detalles Bibliográficos
Autores principales: Bhattacharjee, Sharba, Martin, Ivar
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2503.00241
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866916636388229120
author Bhattacharjee, Sharba
Martin, Ivar
author_facet Bhattacharjee, Sharba
Martin, Ivar
contents We study the retrieval accuracy and capacity of modern Hopfield networks of with two-state (Ising) spins interacting via modified Hebbian $n$-spin interactions. In particular, we consider systems where the interactions deviate from the Hebb rule through additive or multiplicative noise or through clipping or deleting interactions. We find that the capacity scales as $N^{n-1}$ with the number of spins $N$ in all cases, but with a prefactor reduced compared to the Hebbian case. For $n=2$ our results agree with the previously known results for the conventional $n = 2$ Hopfield network.
format Preprint
id arxiv_https___arxiv_org_abs_2503_00241
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Accuracy and capacity of Modern Hopfield networks with synaptic noise
Bhattacharjee, Sharba
Martin, Ivar
Disordered Systems and Neural Networks
We study the retrieval accuracy and capacity of modern Hopfield networks of with two-state (Ising) spins interacting via modified Hebbian $n$-spin interactions. In particular, we consider systems where the interactions deviate from the Hebb rule through additive or multiplicative noise or through clipping or deleting interactions. We find that the capacity scales as $N^{n-1}$ with the number of spins $N$ in all cases, but with a prefactor reduced compared to the Hebbian case. For $n=2$ our results agree with the previously known results for the conventional $n = 2$ Hopfield network.
title Accuracy and capacity of Modern Hopfield networks with synaptic noise
topic Disordered Systems and Neural Networks
url https://arxiv.org/abs/2503.00241