Performance of the DeepTau algorithm for the discrimination of taus against jets, electron, and muons

The performance of the newly developed DeepTau algorithm for the discrimination of taus against jets, electrons, and muons is summarized. The algorithm exploits recent deep neural network multi-classification methods. It outperforms previous discrimination methods significantly in the sense of miss-identification probability of jets, electrons, or muons as taus for given tau identification probability. The agreement of the simulation with the data is good and residual differences are well understood.