Identity Trick#
I think the slides from the MLSS 2018 meeting is the only place that I have encountered anyone actually explicitly mentioning this Identity trick.
Given an integral problem:
I can multiply by an arbitrary distribution which is equivalent to 1.
Then I can regroup and reweight the integral
This results in a different expectation that we initially had
Examples:
Importance Sampling
Manipulate Stochastic gradients
Derive Probability bounds
RL for policy corrections