from neuroclaw52
PyG-based interpretable graph neural network for connectome (fMRI) phenotype prediction, with MPConv message MLP and support for post-hoc edge-mask explanation.
IBGNN (Interpretable Brain Graph Neural Network) provides a PyTorch-Geometric implementation for connectome-based phenotype prediction. It replaces standard GCN message passing with an MLP-based MPConv that computes messages from [x_i, x_j, edge_attr], enabling richer edge-aware representation and supporting downstream edge-mask explanation workflows. NeuroClaw's reimplementation focuses on the encoder and training pipelines for classification and regression tasks on fMRI connectomes.
Use IBGNN when you need to predict subjects' phenotypes (e.g., diagnosis, age, gender) from brain connectivity graphs and want interpretable edge-level insights. Appropriate for research experiments, model comparisons, or when you need a GNN that natively supports attention/edge importance analysis on connectome-style inputs.
Likely used in research or model-execution agents that support Python and PyTorch (OpenClaw/ClawShell, Python-capable agent runtimes).
This skill has not been reviewed by our automated audit pipeline yet.
