Guardado en:
Detalles Bibliográficos
Autores principales: Actor, Jonas A., Gruber, Anthony, Cyr, Eric C.
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2509.04653
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866908613911511040
author Actor, Jonas A.
Gruber, Anthony
Cyr, Eric C.
author_facet Actor, Jonas A.
Gruber, Anthony
Cyr, Eric C.
contents While attention has been empirically shown to improve model performance, it lacks a rigorous mathematical justification. This short paper establishes a novel connection between attention mechanisms and multinomial regression. Specifically, we show that in a fixed multinomial regression setting, optimizing over latent features yields solutions that align with the dynamics induced on features by attention blocks. In other words, the evolution of representations through a transformer can be interpreted as a trajectory that recovers the optimal features for classification.
format Preprint
id arxiv_https___arxiv_org_abs_2509_04653
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Deriving Transformer Architectures as Implicit Multinomial Regression
Actor, Jonas A.
Gruber, Anthony
Cyr, Eric C.
Machine Learning
Artificial Intelligence
Numerical Analysis
While attention has been empirically shown to improve model performance, it lacks a rigorous mathematical justification. This short paper establishes a novel connection between attention mechanisms and multinomial regression. Specifically, we show that in a fixed multinomial regression setting, optimizing over latent features yields solutions that align with the dynamics induced on features by attention blocks. In other words, the evolution of representations through a transformer can be interpreted as a trajectory that recovers the optimal features for classification.
title Deriving Transformer Architectures as Implicit Multinomial Regression
topic Machine Learning
Artificial Intelligence
Numerical Analysis
url https://arxiv.org/abs/2509.04653