Compositional Structures in Neural Embedding and Interaction Decompositions

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Trager, Matthew, Achille, Alessandro, Perera, Pramuditha, Zancato, Luca, Soatto, Stefano
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916320367345664
author Trager, Matthew
Achille, Alessandro
Perera, Pramuditha
Zancato, Luca
Soatto, Stefano
author_facet Trager, Matthew
Achille, Alessandro
Perera, Pramuditha
Zancato, Luca
Soatto, Stefano
contents We describe a basic correspondence between linear algebraic structures within vector embeddings in artificial neural networks and conditional independence constraints on the probability distributions modeled by these networks. Our framework aims to shed light on the emergence of structural patterns in data representations, a phenomenon widely acknowledged but arguably still lacking a solid formal grounding. Specifically, we introduce a characterization of compositional structures in terms of "interaction decompositions," and we establish necessary and sufficient conditions for the presence of such structures within the representations of a model.
format Preprint
id arxiv_https___arxiv_org_abs_2407_08934
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Compositional Structures in Neural Embedding and Interaction Decompositions
Trager, Matthew
Achille, Alessandro
Perera, Pramuditha
Zancato, Luca
Soatto, Stefano
Machine Learning
We describe a basic correspondence between linear algebraic structures within vector embeddings in artificial neural networks and conditional independence constraints on the probability distributions modeled by these networks. Our framework aims to shed light on the emergence of structural patterns in data representations, a phenomenon widely acknowledged but arguably still lacking a solid formal grounding. Specifically, we introduce a characterization of compositional structures in terms of "interaction decompositions," and we establish necessary and sufficient conditions for the presence of such structures within the representations of a model.
title Compositional Structures in Neural Embedding and Interaction Decompositions
topic Machine Learning
url https://arxiv.org/abs/2407.08934