# 语义目标 ERP（TMNRED / ds005383）

核心矩阵协议 · 迁移到新被试

下面这些分数是怎么来的——队列、数据划分、电极、时间窗、每种方法允许学什么、随机猜测能得多少分——都按已发布的协议文件给出。只在本协议内部比较分数。

协议步骤、局限说明、方法备注与数据集署名来自发布数据本身，保持英文原文——它们随数据一起被审核，翻译会让网页与可下载文件不再一致。

## 协议概览

- 任务: Reading · target versus nontarget events
- 评测方式: 迁移到新被试
- 数据集: [TMNRED / ds005383](https://bci.report/zh/datasets/tmnred/)
- 被试: 30
- 数据量: 3,600 epochs · 5 participant-disjoint folds
- 输入: 30 channels · 1-second windows
- 随机水平: 50.0%
- 指标: 平衡准确率（主指标）· 宏平均 F1（次指标）

## 结果

在本协议下运行过的每一种方法，以及已发布结果文件中记录的分数。只在这张表内纵向比较：其他协议在队列、电极、时间窗或随机水平上有所不同。

**平衡准确率，本协议下的每一种方法。** 点为估计值，横线为描述性 95% 区间，虚线为随机水平。

- Spectral ridge: 51.0% (49.6%–52.5%)
- Temporal ridge: 56.3% (54.4%–58.4%)
- LaBraM: 53.2% (51.3%–54.9%)
- CBraMod: 55.8% (54.2%–57.4%)
- EEGNet: 61.4% (58.9%–64.0%)

| 方法 | 训练方式 | 平衡准确率 | 宏平均 F1 | 通道 | 被试 | 评分耗时 |
| --- | --- | --- | --- | --- | --- | --- |
| Spectral ridge — 经典方法 | Supervised fit | 51.0% (49.6%–52.5%) 区间触及随机水平 | 0.506 — Mean across held-out participants | 30 | 30 | 0.9 s |
| Temporal ridge — 经典方法 | Supervised fit | 56.3% (54.4%–58.4%) | 0.559 — Mean across held-out participants | 30 | 30 | 1.2 s |
| [LaBraM](https://bci.report/zh/methods/labram/) — 基础模型 | Frozen encoder + ridge head | 53.2% (51.3%–54.9%) | 0.519 — Mean across held-out participants | 30 | 30 | 3.4 s |
| [CBraMod](https://bci.report/zh/methods/cbramod/) — 基础模型 | Frozen encoder + ridge head | 55.8% (54.2%–57.4%) | 0.554 — Mean across held-out participants | 30 | 30 | 23.5 s |
| [EEGNet](https://bci.report/zh/methods/eegnet/) — 轻量模型 | Scratch · 20 epochs | 61.4% (58.9%–64.0%) | 0.606 — Mean across held-out participants | 30 | 30 | 83.4 s |

评分耗时包括拟合与预测，可能含加速器等待时间，不含数据准备。它只对该配置有效，不是硬件基准。

**未在本协议下运行：** [CSP+LDA](https://bci.report/zh/methods/csp-lda/), [ShallowFBCSPNet](https://bci.report/zh/methods/shallowfbcspnet/), [Deep4Net](https://bci.report/zh/methods/deep4net/), [Standard CCA](https://bci.report/zh/methods/cca/)。这里缺少的方法只是没有在本协议上运行——不是失败。

## 谨慎解读

Balanced event subset with collapsed semantic categories. Natural class prevalence is not preserved. Third-party reading passages and participant metadata are excluded from this website.

### 协议步骤

- 5 participant-disjoint folds. All recordings from a person stay together. Each person contributes to the held-out predictions once.
- units: microvolts; epoch: [event onset, onset + 1.0 s); filtering: none added
- One fixed seed (20260919); no early stopping or test-based tuning. EEGNet trains for 20 epochs per fold. Frozen encoders use training-only standardized ridge heads (alpha 100).
- Labels: nontarget, target.
- Uniform-guessing reference: 50.00%. Scores weight participants equally. Intervals describe participant variation; cross-validation training sets overlap.
- Research subset with fixed deterministic sampling capped at 60 epochs per participant/class.
- Balancing to 60 per class changes the source target/nontarget prevalence.
- Events use numbered target/nontarget variants; the benchmark collapses the suffix only because the trial_type prefix explicitly names the semantic class.

### 稳定性

One fixed seed and training budget; multi-seed sensitivity pending.

### 是否出现在预训练数据中

Unknown unless explicitly documented; no unseen-pretraining claim.

### 方法说明

- One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.

### 模型条款

- **Spectral ridge**：Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used.
- **Temporal ridge**：Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used.
- **LaBraM**：Code/repository: MIT · Checkpoint: committed in that repository; no separate weight terms
- **CBraMod**：Code: MIT · Weights: Apache-2.0 (official model card)
- **EEGNet**：Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used.

## 来源与许可

### TMNRED / ds005383

**署名** Yanru Bai, Qi Tang et al. · TMNRED, A Chinese Language EEG Dataset for Fuzzy Semantic Target Identification in Natural Reading Environments (2025), doi:10.1038/s41597-025-05036-2. OpenNeuro ds005383 v1.0.0.

**许可** [OpenNeuro 元数据为 CC0；配套论文与 GitHub 为 CC BY 4.0 ↗](https://creativecommons.org/licenses/by/4.0/) OpenNeuro metadata says CC0; accompanying publication/GitHub says CC BY 4.0

[数据集记录 ↗](https://doi.org/10.18112/openneuro.ds005383.v1.0.0) · [这个数据集上的全部结果 →](https://bci.report/zh/datasets/tmnred/)

BCI Report 不转发任何记录。这些是 BCI Report 在上述许可下计算的聚合测量；数据归属于署名中的作者。

**许可范围** Personal noncommercial research; aggregate results only

**隐私** 这里只发布队列级聚合结果：不发布任何记录、被试编号或逐人分数。 [完整的审查说明在协议 JSON 中 ↓](https://bci.report/data/semantic-target-protocol.json)

**公开数据登记说明** 有充分的、该研究专属的公开共享证据。聚合指标不会再分发刺激文本与被试元数据；按 CC BY 署名。 Strong study-specific open-sharing evidence. Aggregate metrics avoid redistribution of stimulus text and participant metadata; credit under CC BY.

**权利审查于** 2026-09-20 · 依据 [doi.org ↗](https://doi.org/10.18112/openneuro.ds005383.v1.0.0) · [www.nature.com ↗](https://www.nature.com/articles/s41597-025-05036-2)

## 复现记录

- 协议 ID: `parallel-fixed-subject-folds-v1/ds005383`
- 数据集版本: OpenNeuro snapshot 1.0.0 · ad78f3db430e636595b1b1c08417492f3067a25e
- 硬件: Local Ubuntu / CUDA
- 评分阶段总耗时: 112.4 s · 可能含加速器等待
- 审计记录 SHA-256: `5920974ae9ca3876ce1be82497c139f0783d2ff47173641e329af976441de4c4`
- 汇总 SHA-256: `c5dca7fc30d07a001b8980a4ea65e55ef292f0891a4d32a12262b8ec88389243`
- 协议 SHA-256: `839dfba89cf200aac7a28a7259f45024d006fcd4500170e247fcc1c4e9962c8d`

## 下载

- [结果 · CSV ↓](https://bci.report/data/semantic-target-results.csv)
- [协议 · JSON ↓](https://bci.report/data/semantic-target-protocol.json)
- [核心矩阵 · JSON ↓](https://bci.report/data/experiments.json)

### 其他协议

- [运动想象与静息（ds003810）](https://bci.report/zh/protocols/mi-rest/)
- [空闲与指令（ds005342）](https://bci.report/zh/protocols/idle/)
- [SSVEP · 8 通道（BETA）](https://bci.report/zh/protocols/beta-8ch/)
- [SSVEP · 4 通道（BETA）](https://bci.report/zh/protocols/beta-4ch/)
- [心算与静息（EEGMAT）](https://bci.report/zh/protocols/arithmetic-rest/)
- [P300 目标 ERP（ds006593）](https://bci.report/zh/protocols/p300-target/)
- [睡眠分期（EESM19 scalp subset）](https://bci.report/zh/protocols/sleep-scalp/)

[全部协议 →](https://bci.report/zh/protocols/) · [首页上的核心矩阵 →](https://bci.report/zh/#overview) · [全部数据集 →](https://bci.report/zh/datasets/) · [全部方法 →](https://bci.report/zh/methods/)

---
本文是 https://bci.report/zh/protocols/semantic-target/ 的 Markdown 版本，由已发布的网页生成。数字均为聚合结果；使用条款见 https://bci.report/data-use/（英文）。