Saved in:
| Main Author: | |
|---|---|
| Format: | Recurso digital |
| Language: | |
| Published: |
Zenodo
2025
|
| Online Access: | https://doi.org/10.5281/zenodo.17506760 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Table of Contents:
- <p><span>Self-supervised contrastive learning with </span><span>SimCLR enables 62.1% zero-shot classification accuracy on unseen subjects, significantly outperforming the classical CSP+SVM pipeline (58.3%) in a 9- fold leave-one-subject-out evaluation on the MOABB BNCI2014001 dataset. Paired t-test confirms statistical significance: </span><span>t</span><span>(8) = 6</span><span>.</span><span>12</span><span>, </span><span>p < </span><span>0</span><span>.</span><span>001</span><span>. This work demonstrates that self-supervised pretraining captures subject-invariant motor intention features, eliminating the need for per-user calibration and paving the way for practical, calibration-free BCIs.</span> </p>