Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Liu, Guisong, Gao, Xin, Dresler, Martin, Zhang, Jiansong, Wei, Pengfei
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2605.09905
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866914551747837952
author Liu, Guisong
Gao, Xin
Dresler, Martin
Zhang, Jiansong
Wei, Pengfei
author_facet Liu, Guisong
Gao, Xin
Dresler, Martin
Zhang, Jiansong
Wei, Pengfei
contents Automatic sleep staging commonly adopts Transformers under the assumption that they learn complex long-range dependencies. We challenge this view by revealing a neglected property of sleep sequences: strong local temporal continuity. We show that a randomly initialized Transformer, without any training, substantially improves sleep staging performance and consistently outperforms heuristic smoothing. We formalize this effect via a Random Attention Prior Kernel (RAPK), showing that random self-attention acts as an adaptive smoother by balancing global averaging and content-based similarity while preserving stage transitions. Using two metrics, the Local Smoothness Influence Index (LSII) and the Weighted Transition Entropy (WTE), we provide evidence that most performance gains in Transformer-based sleep staging arise from architectural inductive bias rather than parameter learning. Our results suggest that sleep staging can be effectively addressed with structure-driven smoothing mechanisms rather than complex dependency modeling, enabling more efficient and edge-deployable healthcare systems for large-scale physiological monitoring.
format Preprint
id arxiv_https___arxiv_org_abs_2605_09905
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Rethinking Random Transformers as Adaptive Sequence Smoothers for Sleep Staging
Liu, Guisong
Gao, Xin
Dresler, Martin
Zhang, Jiansong
Wei, Pengfei
Machine Learning
Artificial Intelligence
Automatic sleep staging commonly adopts Transformers under the assumption that they learn complex long-range dependencies. We challenge this view by revealing a neglected property of sleep sequences: strong local temporal continuity. We show that a randomly initialized Transformer, without any training, substantially improves sleep staging performance and consistently outperforms heuristic smoothing. We formalize this effect via a Random Attention Prior Kernel (RAPK), showing that random self-attention acts as an adaptive smoother by balancing global averaging and content-based similarity while preserving stage transitions. Using two metrics, the Local Smoothness Influence Index (LSII) and the Weighted Transition Entropy (WTE), we provide evidence that most performance gains in Transformer-based sleep staging arise from architectural inductive bias rather than parameter learning. Our results suggest that sleep staging can be effectively addressed with structure-driven smoothing mechanisms rather than complex dependency modeling, enabling more efficient and edge-deployable healthcare systems for large-scale physiological monitoring.
title Rethinking Random Transformers as Adaptive Sequence Smoothers for Sleep Staging
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2605.09905