论文标题
在多个行动政策梯度上
On Many-Actions Policy Gradient
论文作者
论文摘要
我们研究每个州有许多动作样本的随机策略梯度（SPG）的方差。我们得出了多个动作的最佳条件，该条件确定了与具有按比例扩展轨迹的单一动力剂相比，SPG何时产生较低的方差。我们提出了基于模型的多功能（MBMA），这是一种在SPG背景下采样的动力学模型的方法。 MBMA解决了与SPG的现有实施相关的问题，并产生较低的偏差和可比较的差异与从模型模拟的推出中估计的SPG相当的差异。我们发现MBMA偏差和方差结构与理论预测的匹配。结果，与无模型，多功能和基于模型的On-Policy On-Policy SPG基准相比，MBMA在一系列连续的动作环境中实现了提高的样品效率和更高的回报。
We study the variance of stochastic policy gradients (SPGs) with many action samples per state. We derive a many-actions optimality condition, which determines when many-actions SPG yields lower variance as compared to a single-action agent with proportionally extended trajectory. We propose Model-Based Many-Actions (MBMA), an approach leveraging dynamics models for many-actions sampling in the context of SPG. MBMA addresses issues associated with existing implementations of many-actions SPG and yields lower bias and comparable variance to SPG estimated from states in model-simulated rollouts. We find that MBMA bias and variance structure matches that predicted by theory. As a result, MBMA achieves improved sample efficiency and higher returns on a range of continuous action environments as compared to model-free, many-actions, and model-based on-policy SPG baselines.
