This is the repository for the paper "LWP-WL: Link weight prediction based on CNNs and Weisfeiler-Lehman algorithm for node ordering"
Please check requirements.txt to check the requisites!
Please after installation run ´python3.8 main.py --help´ to check every hyperparameter and argument values.
Ideally we want to run the Docker installation, so first build the image,
docker image build --tag lwp_wl_pytorch -f LWP-WL_Dockerfile ./
then run the container,
docker run --ipc=host --gpus "device=3" -it --name LWP-WL -v ~/:/code lwp_wl_pytorch bin/bash
The exps.sh file will run standard experiments for LWP-WL, GCNs and Node2Vec (please take into account that optimal parameters for LWP-WL might not be the default ones on each of the dataset).
If you are looking to edit the code he highly suggest to use a binding folder between the machine holding the code and the container.
Otherwise, you can run the code without using Docker whenever you fill the requirements.txt file. The code was tested and published under python3.8.
The exps.sh file will run standard experiments for LWP-WL, GCNs and Node2Vec (please take into account that optimal parameters for LWP-WL might not be the default ones on each of the dataset).