reinforcement-learning

Definition

Definition

The model compounding error typically occurs in sequential decision-making problems where an agent relies on predictions made by a model. Over multiple steps, small errors in these predictions can compound, leading to increasingly inaccurate estimations of the future state or reward.