bnt skillA
bnt is agent-read markdown (skill) from cuhk-aim-group/neurodiscovery: Use this model doc whenever the user wants to run BrainNetworkTransformer for fMRI phenotype prediction, including data loading, training, and evaluation. BNT uses dense FC matrices (no PyG dependency) with DEC pooling + interpretable transformer encoder..
Indexed from public GitHub and served as immutable, content-addressed versions. Install it pinned to an exact SHA-256 with the mdr CLI, and every file is verified against the hash recorded here before it reaches your agent. The deterministic audit below grades the latest version, and the same file always earns the same grade.
What the file says
# BrainNetworkTransformer (BNT) Model Doc ## Overview BrainNetworkTransformer 是一种基于 Transformer 的脑网络分析模型，使用 DEC (Deep Embedded Clustering) 池化实现可解释的节点聚类。 - Paper: Kan et al., 2022, BrainNetworkTransformer - Official code: https://github.com/Wayfear/BrainNetworkTransformer - NeuroClaw reimplementation: `models/bnt/` (无 hydra/omegaconf 依赖，纯 argparse) - Primary input: ROI-level FC 矩阵（dense [N, N]，非 PyG 图格式） - Primary output: phenotype prediction (classification/regression) + attention weights + DEC assignments **Research use only.** --- ## NeuroClaw 实现要点 NeuroClaw 版本对原始 BNT 做了以下关键改动： 1. **去除 hydra/omegaconf 依赖**：原始代码使用 hydra 配置系统，NeuroClaw 改为纯 argparse，降低依赖复杂度。 2. **纯 PyTorch Dataset**：不依赖 PyG，使用标准 `torch.utils.data.Dataset` + `DataLoader`，输入为 dense FC matrix [B, N, N]。 3. **Fisher-z 反变换**：与 BrainGNN 一致，存储的 fc_matrix 是 Fisher-z，加载时用 `torch.tanh()` 还原为 Pearson r，对角线置零。 4. **nhead 自动适配**：当 forward_dim (N + pos_embed_dim) 不能被 nhead 整除时，自动降低 nhead 到最大兼容值。 5. **支持 classification + regression 双任务**：regression 时 nclass=1，输出 raw scalar，用 MSELoss。 6. **可选 T1 GM volume 融合**：`--include-t1` 将 z-scored GM volume 作为额外 1 维拼接到 FC 行末尾（node feature dim 从 N 变为 N+1）。 …
Read the whole file at its exact version.
How to install
mdr add cuhk-aim-group/neurodiscovery/bnt@git:20260515.bfd8daemdr add cuhk-aim-group/neurodiscovery/bnt@sha256:39a4ec57e08c63c9Pin to a label to follow the author's releases, or to a sha256 to freeze the exact bytes forever. Either way the resolved hash is written to mdr.lock, and mdr install reproduces it on any machine.
[](https://markdownregistry.com/a/art_dg7cat74jlzx5hfl)
1 badge views in 30 days
Versions
Audit of the latest version
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (6918 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
cuhk-aim-group/neurodiscovery · 91 stars · license MIT · pushed 2026-09-22 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_dg7cat74jlzx5hfl GET https://markdownregistry.com/api/v1/resolve?ref=cuhk-aim-group/neurodiscovery/bnt GET https://markdownregistry.com/api/v1/blob/39a4ec57e08c63c9fbbc084ec8e4e2f6ccc15ff3274ddeca341703abcad8dcec
Your agent does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.