Python 3.7.9 with following packages

numpy 1.18.5

torch 1.2.0

python3 generate_synthetic_data.py --dataset_id 1

dataset_id: an integer in [1, 12], denoting the index of dataset. There are 12 datasets, see appendix in the paper for details.

python3 train_synthetic.py --dataset_id 1 --time_limit 20000 --algorithm REFS

dataset_id: index of dataset, same with data generation.

algorithm: one of {RAFS, REFS, Brute}. RAFS and REFS are efficient algorithm proposed in the paper, and Brute is a brute-force baseline.

time_limit: the maximum running time in seconds.

The stdout and stderr are redirected to files in following folds:

./log/out/
./log/err/

The file is named with date and time of program execution, e.g. 2021-11-17 16:52:13.526238.