Skip to yearly menu bar Skip to main content


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

Ashley Edwards · Himanshu Sahni · Rosanne Liu · Jane Hung · Ankit Jain · Rui Wang · Adrien Ecoffet · Thomas Miconi · Charles Isbell · Jason Yosinski

Keywords: [ Deep Reinforcement Learning ] [ Reinforcement Learning ] [ Reinforcement Learning - General ]

Abstract: 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

Chat is not available.