arXiv：cs.LG· Bogdan Kozyrskiy, Artem Grachev, Abraham I. Camelo Guerrero·· 4 小时前AI 评分33
EEG BCI 解码的标签揭示在线更新基准测试
Benchmarking Label-Revealed Online Updates for EEG BCI Decoding
AI 导读
一项基准测试在四个数据集（三个运动想象、一个运动解码）上，以时间有序的预序（先测后训）方式对比 CSP 与黎曼协方差两类 EEG BCI 解码流水线。标签揭示的在线更新在两条最大数据流上改善了 14 个模型/数据集组合中的 13 个，相对冻结模型准确率最高提升约 18%。基于 Shapley 的数据估值显示最近时间块的平均价值最高，较旧数据块仍保持正向价值。
正文
Abstract:Electroencephalography (EEG) signals drift over time, which can cause static brain-computer interface (BCI) models to degrade in practice. We present a benchmark for online adaptation and compare two widely used pipeline families, Common Spatial Patterns (CSP) and Riemannian covariance-based methods, under time-ordered prequential (test-then-train) evaluation. We examine (i) which pipelines benefit most from label-revealed updates, (ii) whether controlled forgetting of older data improves robustness, and (iii) how a minimal-calibration cold start compares with starting from a pretrained model. Across four datasets (three motor-imagery datasets and one movement-decoding dataset), label-revealed online updates improve 13 of 14 model/dataset pairs on the two largest streams, with relative accuracy gains of up to about 18% over a frozen model. A Shapley-based data-valuation analysis over temporal blocks assigns the largest mean value to the most recent block in each of the three analyzed datasets, while older blocks retain positive value.
| Subjects: | Machine Learning (cs.LG); Signal Processing (eess.SP) |
| Cite as: | arXiv:2610.07420 [cs.LG] |
| (or arXiv:2610.07420v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.07420 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Artem Grachev [view email]
[v1]
Mon, 5 Oct 2026 21:29:18 UTC (1,083 KB)
来源：arXiv：cs.LG · arxiv.org