You can find the original paper here.
The goal of this project was to analyze how the depth of the neural network (NN) affected the fidelity (measure of reconstruction accuracy) of resolving compressed quantum many-body states using Variational-Autoencoders (VAEs).
That is, we sample the probability density of the state, given by |Ψi (t)|2, and reconstruct this distribution using the generative model and determine fidelity.
.
├── README.md
├── main.py
├── param
│ └── parameters.json
├── results
└── src
├── model
│ ├── hidden_layers.py
│ ├── model.py
│ └── variational_autoencoder.py
└── utils
├── create_dataset.py
├── gen_data
│ ├── gen_hard.py
│ └── gen_random.py
├── get_data.py
└── library.py
usage: main.py [-h] [-v N] [-n N] [--result result/] [--pretrained False]
[--param param/param.json]
Learning Hard Quantum States Using a Variational AutoEncoder
optional arguments:
-h, --help show this help message and exit
-v N Verbosity (0 = all information, else = nothing).
-n N Number of qubits.
--pretrained False Load pretrained model.
--param param/param.json
Parameter file path.
python3 main.py -v 0 -n 18 --param param/parameters.json --pretrained False