True Online Temporal-Difference Learning

By Harm van Seijen, A. Mahmood, Patrick M. Pilarski, Marlos C. Machado, Richard S. Sutton. 2015

105 citations recorded by the source corpus.

The temporal-difference methods TD($\lambda$) and Sarsa($\lambda$) form a core part of modern reinforcement learning. Their appeal comes from their good performance, low computational cost, and their simple interpretation, given by their forward view. Recently, new versions of these methods were introduced, called true online TD($\lambda$) and true online Sarsa($\lambda$), respectively (van Seijen & Sutton, 2014). These new versions maintain an exact equivalence with the forward view at all times, whereas the traditional versions only approximate it for small step-sizes. We hypothesize that these true online methods not only have better theoretical properties, but also dominate the regular methods empirically. In this article, we put this hypothesis to the test by performing an extensive empirical comparison. Specifically, we compare the performance of true online TD($\lambda$)/Sarsa($\lambda$) with regular TD($\lambda$)/Sarsa($\lambda$) on random MRPs, a real-world myoelectric prosthetic arm, and a domain from the Arcade Learning Environment. We use linear function approximation with tabular, binary, and non-binary features. Our results suggest that the true online methods indeed dominate the regular methods. Across all domains/representations the learning speed of the true online methods are often better, but never worse than that of the regular methods. An additional advantage is that no choice between traces has to be made for the true online methods. Besides the empirical results, we provide an in-depth analysis of the theory behind true online temporal-difference learning. In addition, we show that new true online temporal-difference methods can be derived by making changes to the online forward view and then rewriting the update equations.

Subfields: Artificial Intelligence, Biomedical Engineering, Electrical and Electronic Engineering, Reinforcement Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Sergey Levine
2. Pieter Abbeel
3. Yoshua Bengio
4. Shimon Whiteson
5. Ruslan Salakhutdinov
6. Yuxi Li
7. Timothy P. Lillicrap
8. Chelsea Finn

In Biomedical Engineering

1. David Silver
2. Richard S. Sutton
3. Alex Graves
4. Timothy P. Lillicrap
5. Volodymyr Mnih
6. Huizhen Yu

In Electrical and Electronic Engineering

1. Timothy P. Lillicrap
2. Pieter Abbeel
3. Sergey Levine
4. David Silver
5. Koray Kavukcuoglu
6. Volodymyr Mnih
7. Shimon Whiteson
8. Yuxi Li

In Reinforcement Learning

1. Sergey Levine
2. Pieter Abbeel
3. Shimon Whiteson
4. Timothy P. Lillicrap
5. David Silver
6. Yuxi Li
7. R´emi Munos
8. Hado van Hasselt

Citations in the corpus

• Asynchronous Methods for Deep Reinforcement Learning (2016, background): “Yet another, more speculative, direction is to try and combine the recent work on true online temporal difference methods (van Seijen et al., 2015) with non-linear function approximation.”
• Multi-step Off-policy Learning Without Importance Sampling Ratios (2017, methodology): “Equivalence for online updating could also be achieved by following the approach taken by van Seijen et al. (2014, 2016), but we leave that out for simplicity.”