https://github.com/slaclab/neural-representation-sqw
Inelastic neutron scattering parameter estimation using implicit neural representations and automatic differentiation.
https://github.com/slaclab/neural-representation-sqw
Last synced: about 1 year ago
JSON representation
Inelastic neutron scattering parameter estimation using implicit neural representations and automatic differentiation.
- Host: GitHub
- URL: https://github.com/slaclab/neural-representation-sqw
- Owner: slaclab
- License: mit
- Created: 2022-09-27T16:56:50.000Z (over 3 years ago)
- Default Branch: main
- Last Pushed: 2023-12-01T15:20:48.000Z (over 2 years ago)
- Last Synced: 2025-03-28T04:25:40.331Z (about 1 year ago)
- Language: Jupyter Notebook
- Homepage:
- Size: 39.8 MB
- Stars: 3
- Watchers: 8
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
## Capturing dynamical correlations using implicit neural representations
---
## Installation
1) Make a new local folder and clone the repository
```
git clone https://github.com/src47/neural-representation-sqw.git
```
2) Install requirements
```
pip install -r requirements.txt
```
3) Make sure this repo directory is on the PYTHONPATH:
```bash
$ source shell/add_pwd_to_pythonpath.sh
```
## Directory Structure
**data_experimental**
This directory contains experiment S(q,w) measurements for both paths reported in the manuscript (with and without background subtraction). It also contains the accompanying energy and momentum coordinates.
**data_simulation_2023**
Due to the size of the simulation dataset, it is not possible to inclue it directly on GitHub. Please download the simulation data which is publicly available at https://doi.org/10.5281/zenodo.7804447.
**notebooks**
1) test_experimental_data.ipynb: contains code neccesary to optimize the surrogate implict neural model to fit experimental inelastic scattering data.
2) test_low_counts.ipynb: contains code neccesary to fit experimental data as a function of count rate.
**models/siren**
This directory contains a trained SIREN model which acts as a differentiable surrogate for linear spin wave simulations.
## Training Model
To train the SIREN model on simulated excitations from a square lattice, please run:
```bash
$ python3 src/model_training.py --data_path data_simulation_2023/neural_dataset.npz
```
## Citation
If you found this repository useful in your research, please cite:
```bash
@article{chitturi2023capturing,
title={Capturing dynamical correlations using implicit neural representations},
author={Chitturi, Sathya R and Ji, Zhurun and Petsch, Alexander N and Peng, Cheng and Chen, Zhantao and Plumley, Rajan and Dunne, Mike and Mardanya, Sougata and Chowdhury, Sugata and Chen, Hongwei and others},
journal={Nature Communications},
volume={14},
number={1},
pages={5852},
year={2023},
publisher={Nature Publishing Group UK London}
}
```
**Please direct any questions or comments to chitturi@stanford.edu, zhurun@stanford.edu, apetsch@stanford.edu, joshuat@slac.stanford.edu.