combraintf
Use this model doc whenever the user wants to run Com-BrainTF (Community-aware Brain Transformer) for fMRI phenotype prediction. Com-BrainTF uses dense FC matrices with a two-level Transformer (per-community local + global) and DEC pooling. NeuroClaw auto-derives community partitions from atlas naming conventions (Yeo 7-net for Schaefer, lobe-based for AAL).
npx skills add BioTender-max/awesome-bio-agent-skills --skill combraintf --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
# Com-BrainTF Model Doc ## Overview Com-BrainTF (Community-aware Brain Transformer) 是一种针对 fMRI 连接组的两级 Transformer。第一级对每个脑功能社区（如 Yeo 7-network）内的 ROI 独立做 self-attention，并为每个社区维护一个可学习的 CLS token；第二级把所有社区的 CLS + 全部 ROI 节点拼接，再过一个带 DEC 池化的 Transformer，最后展平进 FC head。 - Paper: Bannadabhavi et al., 2023, "Community-Aware Transformer for Autism Prediction in fMRI Connectome"，MICCAI - Official code: https://github.com/ubc-tea/Com-BrainTF - NeuroClaw reimplementation: `models/combraintf/`（去除 hydra/omegaconf 与硬编码 node_clus_map，改为运行时从 atlas 推导） - Primary input: dense FC 矩阵 [B, N, N] - Primary output: phenotype prediction + DEC assignment + per-level attention **Research use only.** --- ## NeuroClaw 实现要点 1. **去除 hydra/omegaconf**：原版用 hydra 配置 + DictConfig，NeuroClaw 改为纯 Python 构造函数，所有参数显式传入。 2. **动态 community partition**：原版从 `node_clus_map.pickle` 加载 Schaefer-400 的固定社区映射；NeuroClaw 在 `data_adapter.py::build_community_ids(atlas)` 里根据 ROI 名自动推导： - `schaefer_*_7net` → Yeo 7-network（Vis/SomMot/DorsAttn/SalVentAttn/Limbic/Cont/Default）+ Unknown 兜底，共 8 组 - `aal_*` / `destrieux` / `dk_*` / `harvard_oxford_*` → 7 lobe + Other = 8 组 - 其他无语义命名的 atlas（cc200/glasser/basc/power/msdl）→ MD5 hash round-robin 8 组兜
- Overview
- NeuroClaw 实现要点
- Quick Start (NeuroClaw 内部)
- 前置条件
- 训练（分类，单 fold 冒烟测试）
- 训练（回归，HCP age）
- 推荐 atlas
- 核心文件
- 模型架构
- 关键训练参数
- 调试经验与注意事项
- NeuroClaw 委托规则
- Reference
python skills/combraintf/scripts/train_reference.py \
What does the combraintf skill do?
Use this model doc whenever the user wants to run Com-BrainTF (Community-aware Brain Transformer) for fMRI phenotype prediction. Com-BrainTF uses dense FC matrices with a two-level Transformer (per-community local + global) and DEC pooling. NeuroClaw auto-derives community partitions from atlas naming conventions (Yeo 7-net for Schaefer, lobe-based for AAL).
How do I install it?
Run `npx skills add BioTender-max/awesome-bio-agent-skills --skill combraintf --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From BioTender-max/awesome-bio-agent-skills, a repository with 135 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.