Priam (Photometric estimation of astrophysical parameters)

Priam is a machine-learning based method that uses two colors (e.g G_Mag - G_BP and G_Mag - G_RP) for AP estimation (i.e. extinction and temperature) of a star. Given that only G_Mag and integrated BP/RP photometry (i.e. G_BP and G_RP) will be released at the early stage in the mission, Priam is useful for preliminary estimations of APs.

To estimate extinction and temperature using colors, we employed a machine learning algorithm, Support Vector Machine (SVM) that are capable of training either a classification model or a regression model using input dataset with known labels or output values.

The above figures show scatter plots of a relation between G_Mag - G_BP versus extinction (temperature) color coded with temperature (extinction) of a training set consisting of about 6500 PHOENIX library stars. As the figures show, temperature and extinction is highly degenerated in the color space. Thus an accuracy for estimation of extinction and temperature using colors is expected to be low.

The above two figures show comparing results between the true APs and the predicted APs. The top panel is a relation between the true APs versus the predicted APs, and the bottom panel is residuals (i.e. true - predicted).