Approximate discounting-free policy evaluation from transient and recurrent states

Vektor Dewanto, Marcus Gallagher

In order to distinguish policies that prescribe good from bad actions in transient states, we need to evaluate the so-called bias of a policy from transient states. However, we obs…