# Time-Efficient Reinforcement Learning with Stochastic Stateful Policies

arXiv: https://arxiv.org/abs/2311.04082 · HTML: https://mlsanity.com/paper/2311.04082/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 4 of its 47 references that are in the map.
- First posted on arXiv 7 Nov 2023.

## Builds on (most cited first, 4 of 4)

- [Real-World Humanoid Locomotion with Reinforcement Learning](https://mlsanity.com/paper/2303.03381/) (2023)
- [Supervised Pretraining Can Learn In-Context Reinforcement Learning](https://mlsanity.com/paper/2306.14892/) (2023)
- [Memory-based control with recurrent neural networks](https://mlsanity.com/paper/1512.04455/) (2015)
- [Visual CPG-RL: Learning Central Pattern Generators for Visually-Guided Quadruped Locomotion](https://mlsanity.com/paper/2212.14400/) (2022)

## Built on by (most cited first, 2 of 2)

- [Real-Time Recurrent Learning using Trace Units in Reinforcement Learning](https://mlsanity.com/paper/2409.01449/) (2024)
- [The Role of Domain Randomization in Training Diffusion Policies for Whole-Body Humanoid Control](https://mlsanity.com/paper/2411.01349/) (2024)

## Abstract (arXiv)

Stateful policies play an important role in reinforcement learning, such as handling partially observable environments, enhancing robustness, or imposing an inductive bias directly into the policy structure. The conventional method for training stateful policies is Backpropagation Through Time (BPTT), which comes with significant drawbacks, such as slow training due to sequential gradient propagation and the occurrence of vanishing or exploding gradients. The gradient is often truncated to address these issues, resulting in a biased policy update. We present a novel approach for training stateful policies by decomposing the latter into a stochastic internal state kernel and a stateless policy, jointly optimized by following the stateful policy gradient. We introduce different versions of the stateful policy gradient theorem, enabling us to easily instantiate stateful variants of popular reinforcement learning and imitation learning algorithms. Furthermore, we provide a theoretical analysis of our new gradient estimator and compare it with BPTT. We evaluate our approach on complex continuous control tasks, e.g., humanoid locomotion, and demonstrate that our gradient estimator scales effectively with task complexity while offering a faster and simpler alternative to BPTT.