Spooky Action at a Distance: Normalization Layers Enable Side-Channel Spatial Communication

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Pfrommer, Samuel, Ma, George, Huang, Yixiao, Sojoudi, Somayeh
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909677314375680
author Pfrommer, Samuel
Ma, George
Huang, Yixiao
Sojoudi, Somayeh
author_facet Pfrommer, Samuel
Ma, George
Huang, Yixiao
Sojoudi, Somayeh
contents This work shows that normalization layers can facilitate a surprising degree of communication across the spatial dimensions of an input tensor. We study a toy localization task with a convolutional architecture and show that normalization layers enable an iterative message passing procedure, allowing information aggregation from well outside the local receptive field. Our results suggest that normalization layers should be employed with caution in applications such as diffusion-based trajectory generation, where maintaining a spatially limited receptive field is crucial.
format Preprint
id arxiv_https___arxiv_org_abs_2507_04709
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Spooky Action at a Distance: Normalization Layers Enable Side-Channel Spatial Communication
Pfrommer, Samuel
Ma, George
Huang, Yixiao
Sojoudi, Somayeh
Machine Learning
This work shows that normalization layers can facilitate a surprising degree of communication across the spatial dimensions of an input tensor. We study a toy localization task with a convolutional architecture and show that normalization layers enable an iterative message passing procedure, allowing information aggregation from well outside the local receptive field. Our results suggest that normalization layers should be employed with caution in applications such as diffusion-based trajectory generation, where maintaining a spatially limited receptive field is crucial.
title Spooky Action at a Distance: Normalization Layers Enable Side-Channel Spatial Communication
topic Machine Learning
url https://arxiv.org/abs/2507.04709