The generalized Bregman divergence is
Where
where the second term is a stabilizer,
Let
Similarly, we can rewrite the second term as
We are given
where
Since
Explicitly, we can frame our problem like this. Given
Since we only need to evaluate