Classification of Astronomical Data (Cherenkov Telescope)
Sören Hader, Fred Hamprecht, Werner Hofmann (Max-Planck-Institute for Nuclear Physics, Heidelberg)
Aim
In cooperation with the Max-Planck-Institut für Kernphysik (Particle Physics and High Energy Astrophysics, Heidelberg, Prof. Hofmann) we want to classify showers triggered by high energy cosmic ray particles with high sensitivity and specificity.
Methods
automatic feature selection
two-stage classification
Examples
Figure 1: Samples from learning set, first 3 rows examples from class "GAMMA", last 3 rows from class "PROTON".
Figure 2: Receiver Operating Characteristic (ROC) from test set, blue line represents the best single classifier (polynomial classifier, 2nd degree) with global and local features. Green line represents the fusion (mean value) of maximum 5 classifiers.
