Deep Learning for Image Sequence Classification of Astronomical Events

Résumé : We propose a new sequential classification model for astronomical objects based on a recurrent convolutional neural network (RCNN) which uses sequences of images as inputs. This approach avoids the computation of light curves or difference images. To the best of our knowledge, this is the first time that sequences of images are used directly for the classification of variable objects in astronomy. In addition, we solve partially the problem of transfer learning from synthetic to real-world images. This is done by transforming synthetic light-curves to images in a realistic way, by taking into account observational conditions and instrumental parameters. This approach allows us to generate datasets to train and test our RCNN model for different astronomical surveys and telescopes. Moreover, using a simulated dataset is faster and more adaptable to different surveys and classification tasks compared to collecting real labeled image sequences. To test the RCNN classifier trained with a synthetic dataset, we used real-world data from the High cadence Transient Survey (HiTS) obtaining an average recall of $87%$ on four classes: supernovae, RR Lyrae, non-variables, and asteroids. We compare the results of our model with those of a light curve classifier, in both simulated and real data. Our RCNN model outperforms the light curve classifier due to the extra information contained on the images. The results obtained encourage us to use and continue developing the proposed method for astronomical alert brokers systems that will process alert streams generated by new telescopes such as the Large Synoptic Survey Telescope and the Zwicky Transient Facility.