automated_gadget_discovery

This repository accompanies the paper "Automated gadget discovery in Science".

Runfiles

Quantum Optics Environment

To reproduce the results for the quantum optics environment, run the command line below which automatically loads the configuration file with the corresponding hyperparameters.

python run_qo_ddqn.py --run {0->9}

Then, run the following run file with the default hyperparameter settings.

pyhton run_qo_clustering.py

After running the files above to generate the data, Table 4 and Table 5 can be reproduce by running the following data analysis file. This file saves the tables as csv-files in the results directory in the corresponding subfolder “analysis-and-plots”.

pyhton run_qo_analysis.py

To try different hyperparameter settings, a copy of the config file exp_0 can be adapted and saved in the same folder under the name exp 1 and then run with the following command:

python run_qo_mcts.py --num_config 1

Quantum Information Environment

To reproduce the results for the quantum information environment, run the command line below which automatically loads the configuration file with the corresponding hyperparameters.

python run_qi_ddqn.py --experiment 'ddqn_cycle_classification' --num_config 3 --run {0->2}

Then, run the following run file with the default hyperparameter settings. The two types of clustering, i.e. by utility and by context, can be reproduced by changing the "clustering_method" argument, in the following line.

python run_qi_clustering.py --clustering_method 'context'

python run_qi_clustering.py --clustering_method 'utility'

After running the files above to generate the data, Fig. 7, Table 8, Table 9 and Table 10 can be reproduced by running the following data analysis file. This file saves the tables as csv-files in the results directory in the corresponding subfolder “analysis-and-plots”.

python run_qi_analysis.py

To visualize circuits in the initialization sets, run the following file:

python run_qi_analysis_init_sets.py

In all files, the main path of the folders can be changed in MAIN_PATH inside the corresponding file.

To try different hyperparameter settings, a copy of the config file exp_3 can be adapted and saved in the same folder under the name exp 4 and then run with the following command:

python run_qi_ddqn.py --num_config 4

Directories

agents

This directory contains the code for the agents: the MCTS and DDQN agent. The code of the DDQN agent is derived from the repository “”.

clustering

This directory contains the code for clustering the gadgets.

configurations

The files containing the parameter configuration to reproduce the results are stored in this directory.

data_mining

This directory contains all the code for the gadget mining.

results

The results produced by any of the runfiles are saved in this directory.