We use the following packages

torch                  1.11.0
geneticalgorithm       1.0.2

First, install the required packages

pip install torch==1.11.0+${CUDA} --extra-index-url https://download.pytorch.org/whl/${CUDA}
pip install geneticalgorithm

We have the following methods

• Baseline methods : greedy-RD / greedy-GA / DC / gurobi
• Ours (UCom2) : ucom2 / ucom2-cpu

Below are the datasets for the robust coloring problem:

• Datasets : collins / gavin / krogan / ppi

Below are the commands for running the baseline methods

python greedy-RD.py --ds [dataset] --c [number_of_colors]
python greedy-GA.py --ds [dataset] --c [number_of_colors]
python DC.py --ds [dataset] --c [number_of_colors]
python gurobi.py --graph_name [dataset] --num_colors [number_of_colors] --timeout_sec 300 --num_seeds 5

Below are the commands for running the proposed method

python ucom2.py --ds [dataset] --c [number_of_colors] --npam [number_of_initial_parameters]
python ucom2-cpu.py --ds [dataset] --c [number_of_colors] --npam [number_of_initial_parameters]

Note: With a larger number of initial parameters, you use more time and memory for better optimization performance