Open Data
Current: /2303.13940
For the code, see https://github.com/tszoldra/attoDNN
Notebooks with examples:
- Dataset and explainability readout
- Simple training and evaluation
- Performance analysis for all models shown in the paper
- Focal averaging example
This database contains:
- 2-dimensional photoelectron momentum distributions resulting from ionization of Argon target by femtosecond laser pulses:
- data_raw/ - results of the simulations and experiments that are in their original format. Details about the data are in the README files.
- data_preprocessed/ - data from data_raw that has a uniform numpy format (same momentum grid) with labels and can be used directly for the machine learning part.
- models/
- models.tar.gz - all 400 pretrained deep convolutional neural network regression models trained in this project (very large)
- QProp_Ar_small_sample.npz_SL_0.0__BayesianVGG16__1000.h5 - one example of a model trained on a small dataset, used in the jupyter notebook example 02_training.ipynb
- analysis_all_model.pkl - python pickle with benchmarks of all trained models. Used in the 03_evaluation_all_models.ipynb notebook.
- explainability.npz - numpy archive with numerical results of the explainability part of the paper. It is used by the 01_dataset_readout.ipynb notebook.
If you use the repository, please cite the paper:
@misc{szoldra2023femtosecond,
title={Femtosecond pulse parameter estimation from photoelectron momenta using machine learning},
author={Tomasz Szołdra and Marcelo F. Ciappina and Nicholas Werby and Philip H. Bucksbaum and Maciej Lewenstein and Jakub Zakrzewski and Andrew S. Maxwell},
year={2023},
eprint={2303.13940},
archivePrefix={arXiv},
primaryClass={physics.atom-ph}
}
[Parent Directory]