1 paper
Wael Hafez, Cameron Reid, Amit Nazeri
Deployed reinforcement learning systems lack a principled runtime reliability theory. We close this gap by introducing Bipredictability, P, a closed form information theoretic metr…