Starting track events in IceCube

The IceCube Neutrino Observatory is a cubic kilometer-sized detector designed to detect neutrinos of astrophysical origin. We summarize an ongoing dataset that will identify starting track neutrino events above 1 TeV over the entire sky. We discuss a method using a boosted decision tree (BDT) to classify and reduce the cosmic ray muon rates from billions per year to ∼1 per year while selecting ∼1000 starting track events per year. Muon tagging with the BDT also improves our sensitivity to astrophysical neutrinos from the southern sky, since it removes atmospheric neutrinos accompanied by muons from parent cosmic ray showers. Next, we introduce a Random Forest to reconstruct the energy of astrophysical neutrino candidates interacting inside the detector volume, with an energy resolution of 25% made possible by identification of the hadronic and muonic components of the neutrino interaction. Finally, the outgoing muon track is used to reconstruct the direction of the astrophysical neutrino candidates with an estimated median angular resolution of 1.6^° at 1 TeV improving to 0.5^° at 1 PeV. This dataset will eventually be used to measure the astrophysical diffuse flux, so we summarize the impact of systematic uncertainties on such a measurement. (paper)