A lot of confusion has been caused over how to use KL Divergence loss function. We need to document it, and this tutorial is an extension of @NRauschmayr's tutorial on loss functions, from which this tutorial will be linked. Shows difference between from_logits as True and False, and discussed various cases where divergence calculated could be wrong/different from true calculation.
A lot of confusion has been caused over how to use KL Divergence loss function. We need to document it, and this tutorial is an extension of @NRauschmayr's tutorial on loss functions, from which this tutorial will be linked. Shows difference between
from_logits
as True and False, and discussed various cases where divergence calculated could be wrong/different from true calculation.