The Statistical Benefits of Quantile Temporal-Difference Learning for Value Estimation
This addresses a performance bottleneck in reinforcement learning for practitioners focused on value estimation, offering a surprising improvement over standard methods.
The paper tackles the problem of temporal-difference-based policy evaluation in reinforcement learning by analyzing quantile temporal-difference learning (QTD), finding that QTD can outperform classical TD learning in predicting mean returns, even when only the mean is of interest, in the tabular setting.
We study the problem of temporal-difference-based policy evaluation in reinforcement learning. In particular, we analyse the use of a distributional reinforcement learning algorithm, quantile temporal-difference learning (QTD), for this task. We reach the surprising conclusion that even if a practitioner has no interest in the return distribution beyond the mean, QTD (which learns predictions about the full distribution of returns) may offer performance superior to approaches such as classical TD learning, which predict only the mean return, even in the tabular setting.