Expected Return Symmetries

Darius Muglich, Johannes Forkel, Elise van der Pol +1

Symmetry is an important inductive bias that can improve model robustness and generalization across many deep learning domains. In multi-agent settings, a priori known symmetries h…