Partially observable MDPs

• The optimal action at time t depends on the entire history of previous observations.

• Instead, a distribution over State(t) suffices.

Previous slide

Next slide

Back to first slide

View graphic version