Scaling Laws for Associative Memories

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Cabannes, Vivien, Dohmatob, Elvis, Bietti, Alberto
Formato: Preprint
Publicado: 2023
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866917594725875712
author Cabannes, Vivien
Dohmatob, Elvis
Bietti, Alberto
author_facet Cabannes, Vivien
Dohmatob, Elvis
Bietti, Alberto
contents Learning arguably involves the discovery and memorization of abstract rules. The aim of this paper is to study associative memory mechanisms. Our model is based on high-dimensional matrices consisting of outer products of embeddings, which relates to the inner layers of transformer language models. We derive precise scaling laws with respect to sample size and parameter size, and discuss the statistical efficiency of different estimators, including optimization-based algorithms. We provide extensive numerical experiments to validate and interpret theoretical results, including fine-grained visualizations of the stored memory associations.
format Preprint
id arxiv_https___arxiv_org_abs_2310_02984
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Scaling Laws for Associative Memories
Cabannes, Vivien
Dohmatob, Elvis
Bietti, Alberto
Machine Learning
Artificial Intelligence
Computation and Language
Neural and Evolutionary Computing
I.2.6; G.1.6
Learning arguably involves the discovery and memorization of abstract rules. The aim of this paper is to study associative memory mechanisms. Our model is based on high-dimensional matrices consisting of outer products of embeddings, which relates to the inner layers of transformer language models. We derive precise scaling laws with respect to sample size and parameter size, and discuss the statistical efficiency of different estimators, including optimization-based algorithms. We provide extensive numerical experiments to validate and interpret theoretical results, including fine-grained visualizations of the stored memory associations.
title Scaling Laws for Associative Memories
topic Machine Learning
Artificial Intelligence
Computation and Language
Neural and Evolutionary Computing
I.2.6; G.1.6
url https://arxiv.org/abs/2310.02984