Von Mises-Fisher Loss for Training Sequence to Sequence Models with Continuous Outputs.

Sachin Kumar, Yulia Tsvetkov

Browse the full ICLR paper archive.