Application of Deep Neural Networks to Event Type Classification in IceCube

The IceCube Neutrino Observatory is able to measure the all-flavor neutrino flux in the energyrange between 100 GeV and several PeV. Due to the different features of the neutrino interactionsand the geometry of the detector, all high-level analyses require a selection of suitable events asa first step. However, presently, no algorithm exists that gives a generic prediction of an event’sunderlying interaction type. One possible solution to this is the use of deep neural networkssimilar to the ones commonly used for 2D image recognition. The classifier that we present hereis based on the modern InceptionResNet architecture and includes multi-task learning in orderto broaden the field of application and increase the overall accuracy of the result. We provide adetailed discussion of the network’s architecture, examine the performance of the classifier forevent type classification and explain possible applications in IceCube.