arXiv：cs.LG（机器学习，全量分类）· Yifan Wang, Haiping Liu, Yang Cui, Wenhao Cai, Shuhang Li, Xiaoyang Huang, Xianyang Liu, Jingyu Sun, Yizheng Sun, Cunhang Fan, Tianming Du, Jiancheng Yang, Zhenhong Li, Yunhao Zhang, Hongpeng Zhou, Jingyuan Sun·· 5 小时前AI 评分41
NeurDuo-EEG：具备持久状态与显式记忆的长序列 EEG 基础模型
NeurDuo-EEG: A Long-Sequence EEG Foundation Model with Persistent State and Explicit Memory
AI 导读
NeurDuo-EEG 是一个带通道级持久记忆的因果 EEG 基础模型，通过多时间尺度记忆管理与选择性检索，实现对连续 EEG 的固定大小状态建模。模型基于 17 个公开数据集、3,955 小时 EEG 预训练，在五个下游基准中的四个取得最佳表现，癫痫检测 AUC-PR 从 0.285 提升至 0.471，Small 变体骨干仅 4.7M 参数，并支持近恒定延迟的流式推理。
正文
Authors:Yifan Wang, Haiping Liu, Yang Cui, Wenhao Cai, Shuhang Li, Xiaoyang Huang, Xianyang Liu, Jingyu Sun, Yizheng Sun, Cunhang Fan, Tianming Du, Jiancheng Yang, Zhenhong Li, Yunhao Zhang, Hongpeng Zhou, Jingyuan Sun
Abstract:Electroencephalography (EEG) is recorded continuously over hours, with relevant dynamics spanning timescales from milliseconds to hours. Most EEG foundation models nevertheless process fixed windows independently, limiting their ability to capture information encoded in long-timescale dynamics. State-space architectures enable persistent recurrent processing, but long-range information remains implicitly compressed in recurrent states. We present NeurDuo-EEG, a causal EEG foundation model with channel-resolved persistent memory. NeurDuo-EEG introduces multi-timescale memory management with learned consolidation and selective retrieval, enabling persistent modelling of continuous EEG with fixed-size state. It is pre-trained on 3,955 hours of EEG from 17 public datasets using multichannel autoregressive prediction of discrete spectral codes. Across three short-window and two long-sequence downstream tasks, NeurDuo-EEG achieves the best performance on four of five benchmarks, including all three short-window tasks and seizure detection, where AUC-PR improves from $0.285$ to $0.471$ over the strongest non-NeurDuo baseline. NeurDuo-EEG also remains competitive on sleep staging and supports efficient streaming inference, with nearly constant per-chunk latency as the available history grows to one hour. Notably, the Small variant achieves this with only 4.7M backbone parameters. These results demonstrate the value of persistent, multi-timescale modelling for both long-sequence and short-window EEG analysis. Our code is available at this https URL.
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2609.38587 [cs.LG] |
| (or arXiv:2609.38587v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2609.38587 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Yifan Wang [view email]
[v1]
Tue, 29 Sep 2026 21:46:11 UTC (1,009 KB)
来源：arXiv：cs.LG（机器学习，全量分类） · arxiv.org