Code for Active Learning for Graph Neural Networks via Node Feature Propagation. The code still needs cleanup. Use with caution.

The package is based on pyg. First run the Anaconda environment installation command conda create -n pyg --file requirement_conda.txt -c pytorch Be sure to set all environment parameters LD_LIBRARY, CPATH, PATH correct (see instructions) before executing the following steps. Then source into pyg environment and install the extra pip packages using pip install -r requirement_pip.txt

run.sh shows a simple example. More options could be found using python active_graph.py --help

Please execute the following commands in sequence: grid.py, dump_csv.py, parse_log.py (get list of metrics), plot.ipynb