QNODE: Learning quantum dynamics using latent neural ODEs

Learning quantum dynamics using latent neural ODEs

Matthew Choi, Daniel Flam-Spepherd, Thi Ha Kyaw, Alán Aspuru-Guzik

https://journals.aps.org/pra/abstract/10.1103/PhysRevA.105.042403

https://arxiv.org/abs/2110.10721

Samples

<p align="center"> <img src="gifs/closed-4.gif" width="250" height="250"> <img src="gifs/closed-6.gif" width="250" height="250"> <img src="gifs/open-7.gif" width="250" height="250"> <img src="gifs/open-10.gif" width="250" height="250"> </p>

Latent Dynamics

<p align="center"> <img src="gifs/latentdynamsclosed.gif" width="250" height="250"> <img src="gifs/latentdynamsopen.gif" width="250" height="250"> </p>

Interpolations

<p align="center"> <img src="gifs/closed interpolate from 235-914.gif" width="250" height="250"> <img src="gifs/open interpolate from 243-269.gif" width="250" height="250"> </p>

Prerequisites

| command | min. version | |:-:|:-:| | torchdiffeq | 0.0.1 | | numpy | 1.17.4 | | Pytorch | 1.4.0 | | QuTip | 4.6.2 | | matplotlib | 3.4.3 | | scikit-learn | 0.23.1 | | imageio | 2.6.1 |

Training Models

run

python3 train.py

To train a model with different hyperparameters: | command | argstype | meaning | |:-:|:-:|:-:| | --seed | int | the torch and numpy random seed | | --epochs | int | numbers of iterations the model will run | | --type | str | either the open or closed dataset | | --obs_dim | int | input dimensions | | --rnn_nhidden | int | rnn layer size | | --nhidden | int | decoder layer size | | --latent_dim | int | latent space size | | --lr | float | learning rate |

Example:

python3 train.py --seed 1 --epochs 5000 --lr 5e-3 --type closed

Generating Results

run

./create_plots.sh

_{^{Note: you might have to run chmod +x create_plots.sh}}