PPG Methods for Parameterized Policy Gradient

Mammal was kind of the prototypical algorithm in the PPG setting. You can add additional parameters to tune other than just the initialization. There's a whole family thing that build on Mammal, and the interloop is consistent between them. And then I guess the only one we haven't really touched on yet is task inference methods. The idea here is a little more nuanced, but meta learning considers a distribution of MDPs, also known as tasks.