Speaker
Description
The analysis of the diffuse neutrino flux with IceCube is usually done using forward-folding binned likelihood fits. The data is binned using energy and angular information. This binning scheme can be improved by employing a neural network optimized summary statistic. For each event, the network predicts a 1D summary statistic which is then binned for the likelihood fit. The network is optimized in such a way that the uncertainties of the fitted signal parameters are improved. In addition to the information utilized in the standard binning, more data can be incorporated as input to the neural network, potentially further enhancing the analysis.
This talk will focus on the implementation of this method for a toy example and the current status of implementing it in the standard analysis framework for diffuse analysis in IceCube.
