centra

Speaker: Dr. Santiago González-Gaitán (CENTRA/IST) for the COIN collaboration

Abstract: 

The modern era of wide-field surveys with powerful cameras in large telescopes opens up the possibility to detect hundreds to thousands of transient phenomena per night. Traditionally these events are classified according to their spectroscopic footprint with the help of spectrographs that require expensive observing time. Such time is currently unavailable for the majority of the massive amount of photometric transients. Machine learning (ML) offers a great possibility to circumvent this problem by providing means to classify objects from photometry directly. For this, one requires a good spectroscopic training sample which is often heavily biased affecting classification. We present here a ML methodology known as "active learning" to obtain an optimal spectroscopic training sample that maximizes the supernova photometric classification.