This is the repo accompanying our paper "Neural Capacitated Clustering" accepted at the 32nd International Joint Conference on Artificial Intelligence (IJCAI23).
The preprint version of our paper can be found here.
install via conda:
conda env create -f requirements.yml
the data.zip is stored via Git LFS. You have to download and unpack it before you can use it.
Simply run the corresponding notebooks to prepare the data and run the evaluation.
In order to create training data, first sample data via ccp_create_data.ipynb. Then create labels via a method of choice using ccp_create_labels.py or vrp_create_labels.py. Finally run the training task runner via run.py. The model is configured via hydra. The configuration files can be found in the config directory. The default config can also be shown via the -h flag
python run.py -h
additional arguments can simply be provided via the command line.