SuperNNova is an open source photometric classification framework. It uses deep learning Recurrent Neural Networks to classify photometric time-series (light-curves).
It does not require feature extraction, is fast and is adaptable to different transients and survey. When trained with large simulations using the observing strategy of a survey, it accurately classifies events and has been applied to data from the Dark Energy Survey (publication) and Zwicky Transient Facility (publication).
With SuperNNova I introduced the use of Bayesian Neural Networks in supernova classification. We aim to use these networks to quantify uncertainties and use them in astrophysical analyses!