Abstract: Context. At GeV energies, the sky is dominated by the interstellar emission from the Galaxy. With limited statistics and spatial resolution, accurately separating point sources is therefore challenging. Aims. Here we present the first application of deep learning based algorithms to automatically detect and classify point sources from gamma-ray data. For concreteness we refer to this approach as AutoSourceID. Methods. To detect point sources, we utilized U-shaped convolutional networks for image segmentation and k-means for source clustering and localization. We also explored the Centroid-Net algorithm, which is designed to find and count objects. Using two algorithms allows for a cross check of the results, while a combination of their results can be used to improve performance. The training data are based on 9.5 years of exposure from The Fermi Large Area Telescope (Fermi-LAT) and we used source properties of active galactic nuclei (AGNs) and pulsars (PSRs) from the fourth Fermi-LAT source catalog in addition to several models of background interstellar emission. The results of the localization algorithm are fed into a classification neural network that is trained to separate the three general source classes (AGNs, PSRs, and FAKE sources). Results. We compared our localization algorithms qualitatively with traditional methods and find them to have similar detection thresholds. We also demonstrate the robustness of our source localization algorithms to modifications in the interstellar emission models, which presents a clear advantage over traditional methods. The classification network is able to discriminate between the three classes with typical accuracy of similar to 70%, as long as balanced data sets are used in classification training. We published online our training data sets and analysis scripts and invite the community to join the data challenge aimed to improve the localization and classification of gamma-ray point sources.

Abstract: A novel method to analyse the spatial distribution of neutrino candidates recorded with the ANTARES neutrino telescope is introduced, searching for an excess of neutrinos in a region of arbitrary size and shape from any direction in the sky. Techniques originating from the domains of machine learning, pattern recognition and image processing are used to purify the sample of neutrino candidates and for the analysis of the obtained skymap. In contrast to a dedicated search for a specific neutrino emission model, this approach is sensitive to a wide range of possible morphologies of potential sources of high-energy neutrino emission. The application of these methods to ANTARES data yields a large-scale excess with a post-trial significance of 2.5 sigma. Applied to public data from IceCube in its IC40 configuration, an excess consistent with the results from ANTARES is observed with a post-trial significance of 2.1 sigma.

Abstract: The results of three different searches for neutrino candidates, associated with the IceCube-170922A event or from the direction of TXS 0506+056, by the ANTARES neutrino telescope, are presented. The first search refers to the online follow-up of the IceCube alert; the second is based on the standard time-integrated method employed by the Collaboration to search for point-like neutrino sources; the third uses information from the IceCube time-dependent analysis that reported bursting activity centered on 2014 December 13, as input for an ANTARES time-dependent analysis. The online follow-up and the time-dependent analysis yield no events related to the source. The time-integrated study performed over a period from 2007 to 2017 fits 1.03 signal events, which corresponds to a p-value of 3.4% (not considering trial factors). Only for two other astrophysical objects in our candidate list has a smaller p-value been found. When considering that 107 sources have been investigated, the post-trial p-value for TXS 0506+056 corresponds to 87%.