Towards a practical measure of interference for reinforcement learning
arXiv:2007.03807
Abstract
Catastrophic interference is common in many network-based learning systems, and many proposals exist for mitigating it. But, before we overcome interference we must understand it better. In this work, we provide a definition of interference for control in reinforcement learning. We systematically evaluate our new measures, by assessing correlation with several measures of learning performance, including stability, sample efficiency, and online and offline control performance across a variety of learning architectures. Our new interference measure allows us to ask novel scientific questions about commonly used deep learning architectures. In particular we show that target network frequency is a dominating factor for interference, and that updates on the last layer result in significantly higher interference than updates internal to the network. This new measure can be expensive to compute; we conclude with motivation for an efficient proxy measure and empirically demonstrate it is correlated with our definition of interference.
18 pages
References in corpus (13)
- Overcoming catastrophic forgetting in neural networks
- Riemannian Walk for Incremental Learning: Understanding Forgetting and Intransigence
- Measuring Catastrophic Forgetting in Neural Networks
- Quantifying Generalization in Reinforcement Learning
- Fantastic Generalization Measures and Where to Find Them
- Assessing Generalization in Deep Reinforcement Learning
- Generalization and Regularization in DQN
- Adapting Auxiliary Losses Using Gradient Similarity
- Towards Characterizing Divergence in Deep Q-Learning
- Stiffness: A New Perspective on Generalization in Neural Networks
- Ray Interference: a Source of Plateaus in Deep Reinforcement Learning
- Toward Understanding Catastrophic Forgetting in Continual Learning
- A Kernel Loss for Solving the Bellman Equation