I have an open question about Prioritized Experience Replay from [Schaul15]. From my experiments, it seems that an equation in the publication is wrong, but maybe I’m overlooking something. I’d appreciate input.

Continue reading →