Reconstructing GeV-scale Neutrinos in IceCube using CNNs

The IceCube Neutrino Observatory instruments 5,160 digital optical modules, which are arrayed over a cubic kilometer deep within the South Pole ice. In the lower center, the modules are more densely configured, which are called the DeepCore subdetector, extending the observable energy threshold down to GeV-scale, where atmospheric neutrino oscillations can be studied. Reconstructing events in the detector is essential in neutrino oscillation analysis. In this talk, I will present the reconstructions of direction, energy, vertex, and particle identifications of GeV-scale events in IceCube by using convolutional neural networks and compare the results to those of the current likelihood-based reconstruction algorithm.