Measuring Interventional Robustness in Reinforcement Learning
This work addresses the issue of robustness in reinforcement learning for researchers and practitioners, but it is incremental as it introduces a new metric without fundamentally changing methods.
The paper tackles the problem of measuring interventional robustness (IR) in reinforcement learning, defining it as the variability in learned policies due to incidental training aspects like data order or exploratory actions, and finds through experiments on eight algorithms in three Atari environments that IR varies with training and algorithm type, and high performance does not guarantee high IR.
Recent work in reinforcement learning has focused on several characteristics of learned policies that go beyond maximizing reward. These properties include fairness, explainability, generalization, and robustness. In this paper, we define interventional robustness (IR), a measure of how much variability is introduced into learned policies by incidental aspects of the training procedure, such as the order of training data or the particular exploratory actions taken by agents. A training procedure has high IR when the agents it produces take very similar actions under intervention, despite variation in these incidental aspects of the training procedure. We develop an intuitive, quantitative measure of IR and calculate it for eight algorithms in three Atari environments across dozens of interventions and states. From these experiments, we find that IR varies with the amount of training and type of algorithm and that high performance does not imply high IR, as one might expect.