Estimating Q(s,s’) with Deep Deterministic Dynamics Gradients

Published in ICML, 2020

In this paper, we introduce a novel form of value function, Q(s, s’), that expresses the utility of transitioning from a state s to a neighboring state s’ and then acting optimally thereafter. In order to derive an optimal policy, we develop a forward dynamics model that learns to make next-state predictions that maximize this value. This formulation decouples actions from values while still learning off-policy. We highlight the benefits of this approach in terms of value function transfer, learning within redundant action spaces, and learning off-policy from state observations generated by sub-optimal or completely random policies. Code and videos are available at

Download paper here

You can cite this work using the BibTeX from ICML, which should be as follows:

  title = 	 {Estimating Q(s,s’) with Deep Deterministic Dynamics Gradients},
  author =       {Edwards, Ashley and Sahni, Himanshu and Liu, Rosanne and Hung, Jane and Jain, Ankit and Wang, Rui and Ecoffet, Adrien and Miconi, Thomas and Isbell, Charles and Yosinski, Jason},
  booktitle = 	 {Proceedings of the 37th International Conference on Machine Learning},
  pages = 	 {2825--2835},
  year = 	 {2020},
  editor = 	 {Hal Daumé III and Aarti Singh},
  volume = 	 {119},
  series = 	 {Proceedings of Machine Learning Research},
  address = 	 {Virtual},
  month = 	 {13--18 Jul},
  publisher =    {PMLR},
  pdf = 	 {},
  url = 	 {}