Learning rate selection via weighted Fisher divergence

Takahiro Onizuka

The general Bayesian approach provides a flexible framework by introducing a loss-based likelihood with a learning rate controlling the relative weight of the likelihood and prior.…