search

SIMEXP / fmri_predict

predicting fmri activaties from connectome
15 stars 11 forks source link

Adding channels for cnn_graph #2

Closed zhangyu2ustc closed 5 years ago

zhangyu2ustc commented 5 years ago

In this project, we used the package by Defferrard: https://github.com/mdeff/cnn_graph

Cite paper: Michaël Defferrard, Xavier Bresson, Pierre Vandergheynst, Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering, Neural Information Processing Systems (NIPS), 2016.

However, in the repo, they only included one channel for each node, which means each node in the graph only has one value as feature. But in our project of fmri/meg decoding, we want to include the time series which represent the BOLD signals. We managed to extend the model to multi-channel and explore other features of gcn

zhangyu2ustc commented 5 years ago

update gcn/lib/model.py line 859:

Build the computational graph.

    self.build_graph((M_0,channel))