Learning Successor Features the Simple Way Raymond Chua Blake A. Richards

In Deep Reinforcement Learning (RL), it is a challenge to learn representations that do not exhibit catastrophic forgetting or interference in non-stationary environments.