IF YOU USE THIS TOOL IN YOUR RESEARCH, PLEASE CITE THE CORRESPONDING PAPER:
https://doi.org/10.1051/0004-6361/201937194
CURRENT INSTALLATION REQUIREMENTS : Python 3, pandas version 0.20 (this specific version is needed, so in a command terminal do: pip install pandas==0.20.3)
"ODUSSEAS.py" is the code we run.
We select the methods by which the reference parameters have been derived, using the setting "reference". This can be: 'photometry' which uses as reference dataset 65 stars with photometric scales of Teff by Casagrande et al (2008) and [Fe/H] by Neves et al (2012), or 'interferometry' (regarded as the new version of ODUSSEAS) which uses as reference dataset 47 stars with interferometry-based Teff by Khata et al (2021) and Rabus et al (2019) and [Fe/H] derived with the method by Neves et al (2012) using the updated parallaxes from Gaia DR3. We can set the regression type using the setting "regression". This can be: 'ridge' (recommended), 'ridgecv', 'linear', 'multitasklasso', 'multitaskelasticnet' We can also choose to do r.v. correction to our spectra if they are shifted, by setting to 'yes' the "do_rv_cor" option.
Input: inside a folder with the path "spectra/newstars/", there should be the fits files of the 1D spectra of the unknown stars. Their filepaths should be written in a text file called "1Dfilelist.dat", and next to them the resolution of each spectrum. See example below:
spectra/newstars/starA.fits 115000
spectra/newstars/starB.fits 94600
spectra/newstars/starC.fits 75000
Output: A text file named "Parameter_Results.dat" is created. It contains the average values of [Fe/H] and Teff after 100 M.L. runs for each star, along with their dispersion, the mean absolute errors of the models that predicted them, the wide error budget (after taking into consideration the intrinsic uncertainties of the reference parameters into the machine learning process), and the machine-learning scores.
Demo set: 1D spectra of stars from 5 different spectrographs with different resolutions and respective HARPS datasets for them are provided to use our tool. For comparison, the reference values of the respective HARPS spectra are the following: Using the scales of Casagrande08 and Neves12: Gl846 = -0.08 & 3682 ; Gl514 = -0.13 & 3574 ; Gl908 = -0.38 & 3587 ; Gl674 = -0.18 & 3284 and for the HARPS star outside the reference HARPS dataset Gl643 = -0.26 & 3102 by Neves et al (2014). Using the scales of Khata21 & Rabus19 and updated Neves12: Gl846 = -0.07 & 3810 ; Gl514 = -0.15 & 3671 ; Gl908 = -0.40 & 3475 ; Gl674 = -0.19 & 3409 ; Gl643 = -0.32 & 3243.
We already provide precomputed pseudo EWs for a range of spectral resolutions used in popular spectrographs. For completeness, the repository also includes the code "HARPS_dataset.py", which can create a library of M dwarfs from our HARPS sample for any resolution we want to work at (the associated fits files are not uploaded). If you wish to create additional libraries or for any other question, please contact us at : alexandros.antoniadis@astro.up.pt
Note: The structure of the repository and associated files should be kept as found, in order for the code to run properly.