Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results.

Antti Tarvainen, Harri Valpola

Browse the full ICLR paper archive.