輸入固定為 32 通道 × 400 samples；比較 SSL full fine-tune、linear probe 與 scratch。
資料快取
| dataset | subject_id | split | kept_windows | excluded_windows | left_trials | right_trials |
|---|---|---|---|---|---|---|
| pretrain | sub16 | train | 1164 | 0 | - | - |
| pretrain | sub16 | val | 125 | 0 | - | - |
| pretrain | sub17 | train | 1146 | 0 | - | - |
| pretrain | sub17 | val | 124 | 0 | - | - |
| pretrain | sub18 | train | 1147 | 0 | - | - |
| pretrain | sub18 | val | 123 | 0 | - | - |
| pretrain | sub19 | train | 1139 | 0 | - | - |
| pretrain | sub19 | val | 123 | 0 | - | - |
| pretrain | sub20 | train | 1170 | 0 | - | - |
| pretrain | sub20 | val | 126 | 0 | - | - |
| pretrain | sub21 | train | 1158 | 0 | - | - |
| pretrain | sub21 | val | 125 | 0 | - | - |
| pretrain | sub22 | train | 1136 | 0 | - | - |
| pretrain | sub22 | val | 123 | 0 | - | - |
| pretrain | sub23 | train | 1154 | 0 | - | - |
| pretrain | sub23 | val | 125 | 0 | - | - |
| pretrain | sub47 | train | 1116 | 0 | - | - |
| pretrain | sub47 | val | 120 | 0 | - | - |
| pretrain | sub49 | train | 1263 | 0 | - | - |
| pretrain | sub49 | val | 136 | 0 | - | - |
| labeled | sub1 | all | 203 | 0 | 100.0 | 103.0 |
| labeled | sub2 | all | 203 | 0 | 102.0 | 101.0 |
| labeled | sub3 | all | 200 | 0 | 100.0 | 100.0 |
| labeled | sub9 | all | 201 | 0 | 100.0 | 101.0 |
| labeled | sub10 | all | 200 | 0 | 100.0 | 100.0 |
| labeled | sub11 | all | 200 | 0 | 100.0 | 100.0 |
| labeled | sub12 | all | 201 | 0 | 100.0 | 101.0 |
| labeled | sub13 | all | 200 | 0 | 100.0 | 100.0 |
| labeled | sub14 | all | 200 | 0 | 100.0 | 100.0 |
| labeled | sub15 | all | 201 | 0 | 100.0 | 101.0 |
CPU smoke test
狀態：passed；遮罩數 [96, 96]；模型參數 3,126,032。
預訓練
遮罩重建
個人化分類結果
群組摘要
| mode | n_subjects | mean_accuracy | std_accuracy | mean_macro_f1 | std_macro_f1 |
|---|---|---|---|---|---|
| ssl_finetune | 10 | 65.97% | 12.64% | 65.22% | 13.30% |
| linear_probe | 10 | 57.90% | 9.05% | 57.43% | 9.30% |
| scratch | 10 | 54.18% | 6.84% | 50.62% | 8.70% |
單一受試者摘要
| subject_id | mode | mean_accuracy | std_accuracy | mean_macro_f1 | std_macro_f1 |
|---|---|---|---|---|---|
| sub1 | linear_probe | 54.06% | 13.42% | 53.54% | 13.35% |
| sub1 | scratch | 49.21% | 6.73% | 40.97% | 8.67% |
| sub1 | ssl_finetune | 68.98% | 7.81% | 68.55% | 8.01% |
| sub10 | linear_probe | 56.50% | 8.77% | 56.40% | 8.76% |
| sub10 | scratch | 52.00% | 3.26% | 47.73% | 7.19% |
| sub10 | ssl_finetune | 79.50% | 11.51% | 79.46% | 11.54% |
| sub11 | linear_probe | 72.50% | 3.95% | 72.40% | 4.06% |
| sub11 | scratch | 51.00% | 4.87% | 48.14% | 7.13% |
| sub11 | ssl_finetune | 86.50% | 5.18% | 86.43% | 5.19% |
| sub12 | linear_probe | 53.30% | 10.86% | 51.71% | 12.78% |
| sub12 | scratch | 54.23% | 5.13% | 50.45% | 10.72% |
| sub12 | ssl_finetune | 62.76% | 11.73% | 61.70% | 13.26% |
| sub13 | linear_probe | 75.50% | 5.12% | 75.35% | 5.31% |
| sub13 | scratch | 69.50% | 8.91% | 68.67% | 9.39% |
| sub13 | ssl_finetune | 78.50% | 9.45% | 78.38% | 9.54% |
| sub14 | linear_probe | 51.00% | 5.48% | 50.77% | 5.60% |
| sub14 | scratch | 50.00% | 7.71% | 49.36% | 7.56% |
| sub14 | ssl_finetune | 52.50% | 8.66% | 51.30% | 8.07% |
| sub15 | linear_probe | 49.24% | 7.27% | 48.87% | 7.13% |
| sub15 | scratch | 50.76% | 3.91% | 44.10% | 8.66% |
| sub15 | ssl_finetune | 47.72% | 11.37% | 45.33% | 14.06% |
| sub2 | linear_probe | 50.76% | 5.23% | 49.50% | 5.07% |
| sub2 | scratch | 47.30% | 2.34% | 41.93% | 5.91% |
| sub2 | ssl_finetune | 57.12% | 8.87% | 56.10% | 9.80% |
| sub3 | linear_probe | 57.00% | 7.98% | 56.72% | 8.15% |
| sub3 | scratch | 62.50% | 6.85% | 61.64% | 8.29% |
| sub3 | ssl_finetune | 68.00% | 6.47% | 67.80% | 6.59% |
| sub9 | linear_probe | 59.17% | 7.39% | 59.04% | 7.26% |
| sub9 | scratch | 55.27% | 8.09% | 53.18% | 8.51% |
| sub9 | ssl_finetune | 58.17% | 10.06% | 57.17% | 10.01% |
成對比較
| first_mode | second_mode | n_subjects | first_mean_accuracy | second_mean_accuracy | mean_paired_difference | bootstrap_ci95_low | bootstrap_ci95_high | wilcoxon_statistic | wilcoxon_p_value |
|---|---|---|---|---|---|---|---|---|---|
| ssl_finetune | scratch | 10 | 65.97% | 54.18% | 11.80% | 5.23% | 19.38% | 3.0 | 0.009766 |
| ssl_finetune | linear_probe | 10 | 65.97% | 57.90% | 8.07% | 3.58% | 12.73% | 4.0 | 0.013672 |