• Run generate_data.py to generate random unitary matrices and control pulses for learning
• Set testing_effectiveness to True in constants_of_experiments.py file.
• Run lstm_as_approximation.py to use generated matrices to train the network and test its efficiency
• Set testing_effectiveness to False in constants_of_experiments.py file.
• Run lstm_as_approximation.py to use the trained network for testing the effect of local disturbances.
• Run printing_results.ipynb to plot results of the experiments.

• generate_data.py - generate data and create directory structure
• get_data.py - functions for loading data from files
• constants_of_experiments.py - this is control panel, with all needed parameters of experiments.
• architecture.py - LSTM architecture and cost functions
• noise_models_and_integration.py - models of quantum systems and related functions
• lstm_as_approximation.py - the main file with experiments.
• printing_results.ipynb - file divided into blocks in which we can plot results of experiment.

The code has been tested with Python 3.6 distributed with Anaconda. The packages utilizes QuTIP is based on TensorFlow library. It also utilizes QuTip for generating control pulses.

M. Ostaszewski, J.A. Miszczak, P. Sadowski, L. Banchi, Approximation of quantum control correction scheme using deep neural networks, https://arxiv.org/abs/1803.05193