The purpose of this repository is to hold a set of tests and ML-based models to predict the temperature of an RR Lyrae-class star based on the wings of its Hydrogen lines (starting with Halpha at 656 nm).
Currently, the temperature of an RR Lyrae star at a snapshot of its phase is obtained by
- capturing a snapshot of its spectrum at a particular point in its phase;
- computing the synthetic spectrum at a given temperature; and
- fitting the synthetic spectrum to the observed one to find a match of the star's temperature.
While this methodology provides good results, it is quite slow: it takes about 5 minutes on an average computer. This is acceptable for small-scale datasets, but we'll need to process very large ones in the future (> 1M stars). In addition, a faster temperature prediction approach may allow us to later develop predictions across the star's phase, improving the field of study of this class of stars.
- Define the input format.
- Get more input data.
- Develop simple estimation models and see how well they behave.
- Estimate the sources of uncertainty and include them on the model.
- Wing selection.
- Uncertainty of the flux measurement.
- Fit error.
- Develop a wing range selection algorithm.
- Play around with different spectrum/flux fitting functions (Chebyshev?).
- Play around with a model that can propagate uncertainties (Metropolis?).
- Look into the possibility of an ML-based de-noising approach. This would broaden the range of the available input data to the noisier regions.