=====
Version 1.0
Artificial Neural-Net compression and fitting of ab initio synthetic spectral grids. The code has the following functionality:
-
Using PyTorch to train an artificial neural-network on individual pixels in a grid of spectra and/or photometry
-
Testing and Validation of a trained ANN to evaluate its precision and calculate the covariance in the predictions
-
Fit observed spectra and/or photometry using trained ANN coupled with Bayesian sampling.
The general algorithm framework for The Payne is given in Ting, Y.-S., Conroy, C., Rix, H.-W., & Cargile, P. 2018, ApJ, submitted. Further details on building ab initio models of synthetic stellar spectra can be found in Ting, Y.-S., Conroy, C., & Rix, H.-W. 2016, ApJ, 826, 83 and Ting, Y.-S., Rix, H.-W., Conroy, C., Ho, A.~Y.~Q., & Lin, J. 2017, ApJL, 849, L9. This code follows the general approach outlined in these papers, with additional changes in the fitting and training procedures. Please cite these references if this code is used for any academic purposes.
The current version of The Payne code is under development and not yet ready for wide distribution. Anyone interested in using this version of the code, please first contact pcargile@cfa.harvard.edu.
The Payne is named in honor of Cecilia Payne-Gaposchkin, one of the great scientist of the 20th century and a pioneering in stellar astrophysicist. In the 1920s she derived the cosmic abundance of the elements from stellar spectra, including determining the composition of the Sun, and demonstrated for the first time the chemical homogeneity of the universe. Cecilia Payne-Gaposchkin achieved two Harvard firsts: she became the first female professor, and the first woman to become department chair.
https://en.wikipedia.org/wiki/Cecilia_Payne-Gaposchkin
- Phillip Cargile (Harvard)
- Yuan-Sen Ting (ANU)
See Authors for a full list of contributors to The Payne.
cd <install_dir>
git clone https://github.com/pacargile/ThePayne.git
cd ThePayne
python -m pip install .
Then in Python
import PayneThe user either need to train a new ANN for an observed spectrum and photometry. Or contact pcargile@cfa.harvard.edu to get the latest C3K based ANN.
The Payne is pure python. See the tutorial for fitting a solar mock with photometric and spectroscopic data. To run demo:
cd demo/
python runPayne.py
The script contains options for fitting solar mock or observed spectra and/or photometry.
Copyright 2018. The Payne is open-source software released under
the MIT License. See the file LICENSE for details.