Measuring and Mitigating Interference in Reinforcement Learning
This work addresses interference issues in reinforcement learning, which is a domain-specific problem for AI researchers and practitioners, but it is incremental as it builds on existing mitigation proposals.
The paper tackled the problem of catastrophic interference in value-based reinforcement learning by introducing a novel measure of interference and showing that it correlates with instability in control performance. They proposed online-aware algorithms that reduce interference and improve stability and performance in classic control environments.
Catastrophic interference is common in many network-based learning systems, and many proposals exist for mitigating it. Before overcoming interference we must understand it better. In this work, we provide a definition and novel measure of interference for value-based reinforcement learning methods such as Fitted Q-Iteration and DQN. We systematically evaluate our measure of interference, showing that it correlates with instability in control performance, across a variety of network architectures. Our new interference measure allows us to ask novel scientific questions about commonly used deep learning architectures and study learning algorithms which mitigate interference. Lastly, we outline a class of algorithms which we call online-aware that are designed to mitigate interference, and show they do reduce interference according to our measure and that they improve stability and performance in several classic control environments.