Generalized End-to-End Loss for Speaker Verification

Li Wan, Quan Wang, Alan Papir +1

In this paper, we propose a new loss function called generalized end-to-end (GE2E) loss, which makes the training of speaker verification models more efficient than our previous tu…