I wonder how similar it is to using KLD to prior label distribution as a secondary term, and decrease its weight during training?
SoftTarget, an interesting new regularization mechanism for neural networks: http://arxiv.org/abs/1609.06693
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
does it really work? I wish the author would just public validation accuracy info in addition to loss.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.