Wednesday, February 1, 2012

1201.6676 (Emille E. O. Ishida et al.)

Kernel PCA for type Ia supernovae photometric classification    [PDF]

Emille E. O. Ishida, Rafael S. de Souza
The problem of photometric identification will be extremely important for large surveys in the next decade. In this work, we propose the use of KPCA combined with k = 1 nearest neighbor algorithm (KPCA+1NN) as a framework for SNe photometric classification. The method does not rely on information about redshift or local enviromental variables, so it is less sensitive to bias than its template fitting counterparts. We applied the method to $\approx$ 20000 SNe light curve released after the \textit{Supernova Photometric Classification Challenge} (SNPCC). Results for the photometric sample achieved up to 89% efficiency (eff), 97% purity (pur), 96% successful classification (SC) rates and figure of merit (FoM) of 0.79 (SNR$\geq$5). If we impose no SNR cuts, we obtain up to 64% eff, 43% pur, 46% SC and FoM of 0.10. We also present the classification results using only pre-maximum epoches, obtaining 80% eff, 73% pur, 84% SC and FoM of 0.32 (SNR$\geq$5). Comparing the performance of our classifier with MLCS2k2 fit probability, we demonstrate that KPCA+1NN is able to improve the classification results of MLCS2k up to $\approx 15%$ without the need of redshift information. Results are sensitive to the information contained in each light curve, as a consequence, higher quality the data points lead to higher successfull classification rates. The method is flexible enough to be applied to other astrophysical transients, as long as a training and a template sample are provided.
View original: http://arxiv.org/abs/1201.6676

No comments:

Post a Comment