Pseudorehearsal in value function approximation
This addresses forgetting in non-stationary RL environments, but it is incremental as it applies known techniques to a simple task.
The paper tackled catastrophic forgetting in reinforcement learning by comparing pseudorehearsal methods for Q-learning in a pole balancing task, finding that it assists learning with proper parameter initialization.
Catastrophic forgetting is of special importance in reinforcement learning, as the data distribution is generally non-stationary over time. We study and compare several pseudorehearsal approaches for Q-learning with function approximation in a pole balancing task. We have found that pseudorehearsal seems to assist learning even in such very simple problems, given proper initialization of the rehearsal parameters.