Deep autoregressive density nets vs neural ensembles for model-based offline reinforcement learning
This work addresses offline reinforcement learning for robotics and control systems, offering a more efficient alternative to ensembles, though it is incremental as it builds on existing model-based approaches.
The paper tackles the problem of model error exploitation in offline reinforcement learning by showing that a single well-calibrated autoregressive model outperforms ensemble methods on the D4RL benchmark, achieving better performance without relying on ensembles.
We consider the problem of offline reinforcement learning where only a set of system transitions is made available for policy optimization. Following recent advances in the field, we consider a model-based reinforcement learning algorithm that infers the system dynamics from the available data and performs policy optimization on imaginary model rollouts. This approach is vulnerable to exploiting model errors which can lead to catastrophic failures on the real system. The standard solution is to rely on ensembles for uncertainty heuristics and to avoid exploiting the model where it is too uncertain. We challenge the popular belief that we must resort to ensembles by showing that better performance can be obtained with a single well-calibrated autoregressive model on the D4RL benchmark. We also analyze static metrics of model-learning and conclude on the important model properties for the final performance of the agent.