부산대학교 도서관

Deep convolutional neural networks (DCNs) are a promising machine learning technique to reconstruct events recorded by imaging atmospheric Cherenkov telescopes (IACTs), but require optimization to reach full performance. One of the most pressing challenges is processing raw images captured by cameras made of hexagonal lattices of photo-multipliers, a common layout among IACT cameras which topologically differs from the square lattices conventionally expected, as their input data, by DCN models. Strategies directed to tackle this challenge range from the conversion of the hexagonal lattices onto square lattices by means of oversampling or interpolation to the implementation of hexagonal convolutional kernels. In this contribution we present a comparison of several of those strategies, using DCN models trained on simulated IACT data.
Comment: 8 pages, 3 figures, Proceedings of the ICRC 2019 PoS(ICRC2019)753