BOSSNet

Credit: Logan Sizemore, Diego Llanes, Indie Cowan, Dylan Huson

BOSSNet is a pipeline for predicting stellar parameters, including effective temperature, surface gravity, metallicity, and radial velocity. It supports various astronomical surveys and allows users to compute predictions with optional uncertainty estimates.

BOSS Net Paper

GAIA Net Paper

Supported Surveys

BOSSNet supports the following data sources:

• BOSS (Baryon Oscillation Spectroscopic Survey)
• APOGEE (Apache Point Observatory Galactic Evolution Experiment)
• LAMOST (Large Sky Area Multi-Object Fiber Spectroscopic Telescope), including:
  • LAMOST DR7
  • LAMOST DR8
• Gaia (Global astrometric and spectroscopic survey), including:
  • Gaia XP (Extended Photometric Survey)
  • Gaia RVS (Radial Velocity Spectrometer)

Input File Requirements

• BOSS, APOGEE, and LAMOST (DR7/DR8): Provide a plain text file containing a list of paths to the FITS files.
• Gaia: Provide an ECSV file in the format found at:
  • Gaia XP Continuous Mean Spectrum
  • Gaia RVS Mean Spectrum

Usage

To execute the pipeline, use the following command structure:

python bossnet <spectra_paths> -d <data_source> -o <output_file> -u <num_uncertainty_draws> -b <batch_size> -w <num_workers> -v

Positional Arguments

• spectra_paths: Path to a plain text file containing paths to the spectra used for predictions.

Optional Arguments

• -h, --help: Display help information and exit.
• -d, --data_source: Source of data. Default is boss. Available options are:
  • boss
  • apogee
  • lamost_dr7
  • lamost_dr8
  • gaia_xp
  • gaia_rvs
• -o, --output_file: Path to the file where predictions will be saved. If not provided, the predictions will go to standard out.
• -u, --num_uncertainty_draws: Number of realizations to sample from the predictive distribution for uncertainty calculation. Default is 0.
• -b, --batch_size: Batch size for the data loader. Adjust this based on system memory availability.
• -w, --num_workers: Number of workers for the data loader to utilize for parallel data loading.
• -v, --verbose: Print progress messages to the console for better visibility of the execution process.

Example Commands

1. Predict stellar parameters for BOSS data and save results to a file:

   python bossnet spectra_paths.txt -d boss -o predictions.txt -v

2. Include uncertainty estimation with 100 draws:

   python bossnet spectra_paths.txt -d apogee -o predictions_with_uncertainties.txt -u 100 -v

3. Use Gaia XP data with specific batch size and workers:

   python bossnet gaia_spectra.csv.gz -d gaia_xp -b 32 -w 4 -o gaia_predictions.txt

Output

The output file (if specified) will contain predictions for each spectrum provided in the input file, including calculated stellar parameters and optional uncertainty estimates.