| |
 |
| Paper: |
Classification and Astrophysical Parameter Estimation from Gaia: Design and First Results |
| Volume: |
376, Astronomical Data Analysis Software and Systems XVI |
| Page: |
421 |
| Authors: |
Tiede, C.; Holmberg, J.; Bailer-Jones, C.A.L. |
| Abstract: |
Gaia is the next generation astrometric mission now under construction by ESA to study the composition and evolution of our Galaxy. As Gaia performs real time detection, the intrinsic properties of most of the objects it observes will not be known a priori. An integral part of the Gaia data processing is therefore to classify all observed objects and to determine their intrinsic astrophysical parameters. For this purpose Gaia is equipped with a low dispersion slitless spectrophotometer covering the wavelength range 330–680 nm, (blue photometer, or BP) and 640–1000 nm (red photometer, or RP), although the classification work will also exploit the astrometry and onboard high-resolution spectroscopy (used primarily to derive radial velocities). We describe the approach being taken by the data processing consortium to classify and parameterize the data. This can be split into discrete classification, continuous parameter estimation and an unsupervised analysis for knowledge discovery. In our initial test, the discrete classification is carried out by applying support vector machines. The results show a good accuracy at classifying most of the objects. The parameter estimation is done via a nearest neighbor and a support vector machine approach. Both results show a good correlation between true and estimated values for those parameters which have a strong effect on the spectra. The correlation of those parameters to which the spectra have low sensitivity is quite small. In the future we will improve both methods by domain-specific modifications. |
|
|
|
|
|
