Diffusion Policies as an Expressive Policy Class for Offline Reinforcement Learning
By Zhendong Wang, Jonathan J. Hunt, Mingyuan Zhou. International Conference on Learning Representations, 2022
677 citations recorded by the source corpus.
Offline reinforcement learning (RL), which aims to learn an optimal policy using a previously collected static dataset, is an important paradigm of RL. Standard RL methods often perform poorly in this regime due to the function approximation errors on out-of-distribution actions. While a variety of regularization methods have been proposed to mitigate this issue, they are often constrained by policy classes with limited expressiveness that can lead to highly suboptimal solutions. In this paper, we propose representing the policy as a diffusion model, a recent class of highly-expressive deep generative models. We introduce Diffusion Q-learning (Diffusion-QL) that utilizes a conditional diffusion model to represent the policy. In our approach, we learn an action-value function and we add a term maximizing action-values into the training loss of the conditional diffusion model, which results in a loss that seeks optimal actions that are near the behavior policy. We show the expressiveness of the diffusion model-based policy, and the coupling of the behavior cloning and policy improvement under the diffusion model both contribute to the outstanding performance of Diffusion-QL. We illustrate the superiority of our method compared to prior works in a simple 2D bandit example with a multimodal behavior policy. We then show that our method can achieve state-of-the-art performance on the majority of the D4RL benchmark tasks.
Subfields: Artificial Intelligence
View Paper: https://www.semanticscholar.org/paper/2cbea7615ebecea2c414d8fbad47d5d258a5c3b4
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
Citations in the corpus
- 3D Diffuser Actor: Policy Diffusion with 3D Scene Representations (2024, positive): “They have been used for modeling state-conditioned action distributions in imitation learning [62, 77, 54, 78, 60, 51] from low-dimensional input, as well as from visual sensory input, and show both better mode coverage and higher fidelity in action prediction than alternatives.”
- Directly Fine-Tuning Diffusion Models on Differentiable Rewards (2023, critical): “Unfortunately, we found this method to be unstable and ultimately not learn effectively, possibly because it relies on the assumption that ∇ x r ( x 0 , c ) ≈ ∇ x r (ˆ x 0 , c ) , which may not be true in practice.”
- NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration (2023, background): “State-or observation-conditioned diffusion models of action [30, 31] are particularly powerful, since they enable modeling complex action distributions without the cost and added complexity of inferring future states/observations.”
- IDQL: Implicit Q-Learning as an Actor-Critic Method with Diffusion Policies (2023, neutral): “Diffusion Q-learning (DQL) [44] incorporates diffusion to parameterize the actor in a TD3+BC-style algorithm [12].”
- Training Diffusion Models with Reinforcement Learning (2023, positive): “…in applications including image and video synthesis (Ramesh et al., 2021; Saharia et al., 2022; Ho et al., 2022), drug and material design (Xu et al., 2021; Xie et al., 2021; Schneuing et al., 2022), and continuous control (Janner et al., 2022; Wang et al., 2022; Hansen-Estruch et al., 2023).”
- Learning Universal Policies via Text-Guided Video Generation (2023, methodology): “Diffusion models have recently been applied to different decision making problems (Janner et al., 2022; Ajay et al., 2022; Wang et al., 2022; Zhang et al., 2022a; Urain et al., 2022; Zhong et al., 2022).”
- Diffusion policy: Visuomotor policy learning via action diffusion (2023, positive): “In the context of Reinforcement Learning, Wang et al. (2022) use diffusion model for policy representation and regularization with state-based observations.”
- Is Conditional Generative Modeling all you need for Decision-Making? (2022, methodology): “Recent works have also used diffusion models to imitate human behavior (Pearce et al., 2023) and to parameterize policy in offline RL (Wang et al., 2022).”