Dataset for "Reinforcement Learning of Quantum Circuit Architectures for Molecular Potential Energy Curves"

This dataset contains the raw simulation data used in the paper “Reinforcement Learning of Quantum Circuit Architectures for Molecular Potential Energy Curves”. All data were generated using our RL-based framework, which is available in the accompanying GitHub repository. LiH_4q_2.2Numerical results for four-qubit lithium hydride at a single bond distance of 2.2 Å. LiH_4q_1.0_4.0Numerical results for four-qubit lithium hydride used to generate the potential energy curve (PEC) over the bond-distance interval 1.0–4.0 Å. LiH_6q_2.2Numerical results for six-qubit lithium hydride at a single bond distance of 2.2 Å. LiH_6q_1.0_4.0Numerical results for six-qubit lithium hydride used to generate the PEC over the bond-length interval 1.0–4.0 Å. H4_q8_1.5Numerical results for eight-qubit H4 at a single bond distance of 1.5 Å. H4_q8_0.55_1.55Numerical results for eight-qubit H4 used to generate the PEC over the bond-length interval 0.55–1.55 Å. JSPNumerical results for the job shop scheduling (JSP) setup. For each setup, the corresponding folder contains a config.py file that was used to run this setup within our RL-VQE framework, as well as a molecular_data.npy file, which stores the molecular Hamiltonian, ground-state energy and Hartree–Fock energy for the system. Each setup folder further contains 12 subfolders named training_session_x, which store the results of the x-th independent training run for this setup. Each training_session_x folder includes: actor_model_x.pt: the learned parameters of the neural network (actor) after training. rl_quantities_x.npy: reinforcement-learning-specific quantities such as actor and critic losses, entropy, and related diagnostics. results.npy: per-episode results, including the obtained energy, the constructed circuit, and the corresponding reward. unseen_predictions_x.npy (PEC setups only): energy evaluations of the trained policy on unseen bond distances For further implementation details, please refer to the associated GitHub repository.