This repository provides data and supplementary material to the paper Bidirectional deep learning of polarization transfer in liquid crystals with application to quantum state preparation, Physical Review Applied 17, 054042 (2022), by Dominik Vašinka, Martin Bielak, Michal Neset, and Miroslav Ježek.
Auxiliary_models folder
- contains neural networks for evaluating the Fig. 2 (c): Dependence on the number of trainable parameters
Compound_model.h5
- a file containing a deep neural network built in Keras library with Tensorflow backend
- the network coresponds to the compound model from section "Inverse and compound model"
Data_files folder
- contains files required by corresponding jupyter notebooks
Dataset_size_scaling.ipynb
- a Jupyter Notebook for reproducting the Fig. 2 (a): The infidelity dependence on the number of samples in the training set
- requires loading of the following files contained within "Data_files" folder:
- DNN_all_fidelity_val_tensor_part_1.npy, DNN_all_fidelity_val_tensor_part_2.npy
- DNN_all_fidelity_test_tensor_part_1.npy, DNN_all_fidelity_test_tensor_part_2.npy
- LI_all_fidelity_val_tensor_part_1.npy, LI_all_fidelity_val_tensor_part_2.npy
- LI_all_fidelity_test_tensor_part_1.npy, LI_all_fidelity_test_tensor_part_2.npy
- RBF_all_fidelity_val_tensor_part_1.npy, RBF_all_fidelity_val_tensor_part_2.npy
- RBF_all_fidelity_test_tensor_part_1.npy, RBF_all_fidelity_test_tensor_part_2.npy
Direct_model.h5
- a file containing a deep neural network built in Keras library with Tensorflow backend
- the neural network coresponds to the direct model from section "Direct models and dataset-size scaling"
Evaluation_of_DNN.ipynb
- a Jupyter Notebook for reproducing the infidelity results of direct and compound deep learning models
- requires loading of models "Direct_model.h5" and "Compound_model.h5", together with following files from "Data_files" folder:
- training_voltages.npy, validation_voltages.npy, test_voltages.npy
- training_rhos.npy, validation_rhos.npy, test_rhos.npy
Evaluation_of_interpolation_methods.ipynb
- a Jupyter Notebook for reproducing the infidelity results of linear and radial basis function interpolation models
- requires loading of the following files from "Data_files" folder:
- training_voltages.npy, validation_voltages.npy, test_voltages.npy
- training_rhos.npy, validation_rhos.npy, test_rhos.npy
Logo_projection.ipynb
- a Jupyter Notebook for reproducting the Fig. 5: Palacký University logos on the Bloch sphere
- requires loading of the "Compound_model.h5" file, together with the following files from "Data_files" folder:
- logo_points_cartesian.npy
- weak_signal_heralded_stokes.npy
- entangled_state_density_matrix_MN.npy
All the files were prepared and tested with Python 3.9.6, Keras 2.4.3, and Tensorflow 2.5.0