| |
 |
| Paper: |
Instance-Based Machine Learning Methods for the Prediction of Stellar Atmospheric Parameters |
| Volume: |
216, Astronomical Data Analysis Software and Systems IX |
| Page: |
611 |
| Authors: |
Fuentes, O.; Gulati, R. K. |
| Abstract: |
We have performed an experimental comparison of several instance-based machine learning algorithms applied to the problem of automatically estimating stellar atmospheric parameters from their spectral indices. We have implemented nearest-neighbors and locally weighted regression algorithms, introducing also a dimensionality reduction preprocessing stage using principal component analysis. Our experimental results show that these algorithms are capable of predicting effective temperature, surface gravity and metallicity quickly and accurately. We observe that dimensionality reduction, besides significantly reducing the computation time, can also improve the accuracy of the resulting predictions. We envisage the use of such methods for large spectroscopic surveys currently in progress. |
|
|
|
|
|
