Primal-Dual Neural Algorithmic Reasoning

Installation

1. Create the environment using the environment.yml file:

   conda env create -f environment.yml

2. Activate the environment:

   conda activate pdnar

Usage

To run the experiments for the three NP-hard problems, use the following commands:

1. Minimum Vertex Cover (MVC)

   python main.py data.algorithm="vertex_cover"

2. Minimum Set Cover (MSC)

   python main.py data.algorithm="set_cover"

3. Minimum Hitting Set (MHS)

   python main.py data.algorithm="hitting_set" model.model.eps=True 

• Note that model.model.eps controls whether the uniform increase rule in used.

Optional parameters

Parameter	Description	Type	Default Value
seed	Random seed.	int	0
wandb_use	Whether to use wandb.	bool	False
inference_only	Whether to only perform inference.	bool	False
checkpoint	Path of pretrained model for inference only.	str	null