Policy gradient methods are an appealing approach in reinforcement learning because they directly optimize the cumulative reward and can straightforwardly be used with nonlinear function approximators such as neural networks. The two main challenges are the large number of samples typically required, and the difficulty of obtaining stable and steady improvement despite the nonstationarity of the incoming data. We address the first challenge by using value functions to substantially reduce the variance of policy gradient estimates at the cost of some bias, with an exponentially-weighted estimator of the advantage function that is analogous to TD(λ). We address the second challenge by using trust region optimization procedure for both the policy and the value function, which are represented by neural networks. Our approach yields strong empirical results on highly challenging 3D locomotion tasks, learning running gaits for bipedal and quadrupedal simulated robots, and learning a policy for getting the biped to stand up from starting out lying on the ground. In contrast to a body of prior work that uses hand-crafted policy representations, our neural network policies map directly from raw kinematics to joint torques. Our algorithm is fully model-free, and the amount of simulated experience required for the learning tasks on 3D bipeds corresponds to 1-2 weeks of real time.

We describe an iterative procedure for optimizing policies, with guaranteed monotonic improvement. By making several approximations to the theoretically-justified procedure, we develop a practical algorithm, called Trust Region Policy Optimization (TRPO). This algorithm is similar to natural policy gradient methods and is effective for optimizing large nonlinear policies such as neural networks. Our experiments demonstrate its robust performance on a wide variety of tasks: learning simulated robotic swimming, hopping, and walking gaits; and playing Atari games using images of the screen as input. Despite its approximations that deviate from the theory, TRPO tends to give monotonic improvement, with little tuning of hyperparameters.

政策梯度定理（Sutton等，2000）规定了目标政策下的累积折扣国家分配以近似梯度。实际上，基于该定理的大多数算法都打破了这一假设，引入了分布转移，该分配转移可能导致逆转溶液的收敛性。在本文中，我们提出了一种新的方法，可以从开始状态重建政策梯度，而无需采取特定的采样策略。可以根据梯度评论家来简化此形式的策略梯度计算，由于梯度的新钟声方程式，可以递归估算。通过使用来自差异数据流的梯度评论家的时间差异更新，我们开发了第一个以无模型方式避开分布变化问题的估计器。我们证明，在某些可实现的条件下，无论采样策略如何，我们的估计器都是公正的。我们从经验上表明，我们的技术在存在非政策样品的情况下实现了卓越的偏见变化权衡和性能。

In this paper we consider deterministic policy gradient algorithms for reinforcement learning with continuous actions. The deterministic policy gradient has a particularly appealing form: it is the expected gradient of the action-value function. This simple form means that the deterministic policy gradient can be estimated much more efficiently than the usual stochastic policy gradient. To ensure adequate exploration, we introduce an off-policy actor-critic algorithm that learns a deterministic target policy from an exploratory behaviour policy. We demonstrate that deterministic policy gradient algorithms can significantly outperform their stochastic counterparts in high-dimensional action spaces.

Model-free deep reinforcement learning (RL) algorithms have been demonstrated on a range of challenging decision making and control tasks. However, these methods typically suffer from two major challenges: very high sample complexity and brittle convergence properties, which necessitate meticulous hyperparameter tuning. Both of these challenges severely limit the applicability of such methods to complex, real-world domains. In this paper, we propose soft actor-critic, an offpolicy actor-critic deep RL algorithm based on the maximum entropy reinforcement learning framework. In this framework, the actor aims to maximize expected reward while also maximizing entropy. That is, to succeed at the task while acting as randomly as possible. Prior deep RL methods based on this framework have been formulated as Q-learning methods. By combining off-policy updates with a stable stochastic actor-critic formulation, our method achieves state-of-the-art performance on a range of continuous control benchmark tasks, outperforming prior on-policy and off-policy methods. Furthermore, we demonstrate that, in contrast to other off-policy algorithms, our approach is very stable, achieving very similar performance across different random seeds.

由于策略梯度定理导致的策略设置存在各种理论上 - 声音策略梯度算法，其为梯度提供了简化的形式。然而，由于存在多重目标和缺乏明确的脱助政策政策梯度定理，截止策略设置不太明确。在这项工作中，我们将这些目标统一到一个违规目标，并为此统一目标提供了政策梯度定理。推导涉及强调的权重和利息职能。我们显示多种策略来近似梯度，以识别权重（ACE）称为Actor评论家的算法。我们证明了以前（半梯度）脱离政策演员 - 评论家 - 特别是offpac和DPG - 收敛到错误的解决方案，而Ace找到最佳解决方案。我们还强调为什么这些半梯度方法仍然可以在实践中表现良好，表明ace中的方差策略。我们经验研究了两个经典控制环境的若干ACE变体和基于图像的环境，旨在说明每个梯度近似的权衡。我们发现，通过直接逼近强调权重，ACE在所有测试的所有设置中执行或优于offpac。

大多数加固学习算法优化了折扣标准，这些标准是有益的，可以加速收敛并降低估计的方差。虽然折扣标准适用于诸如财务相关问题的某些任务，但许多工程问题同样对待未来的奖励，并更喜欢长期的平均标准。在本文中，我们研究了长期平均标准的强化学习问题。首先，我们在折扣和平均标准中制定统一的信任区域理论，并在扰动分析（PA）理论中导出信托区域内的新颖性能。其次，我们提出了一种名为平均策略优化（APO）的实用算法，其提高了名为平均值约束的新颖技术的值估计。最后，实验在连续控制环境Mujoco中进行。在大多数任务中，APO比折扣PPO更好，这表明了我们方法的有效性。我们的工作提供了统一的信任地区方法，包括折扣和平均标准，这可能会补充折扣目标超出了钢筋学习的框架。

Softmax政策的政策梯度（PG）估计与子最佳饱和初始化无效，当密度集中在次良动作时发生。从策略初始化或策略已经收敛后发生的环境的突然变化可能会出现次优策略饱和度，并且SoftMax PG估计器需要大量更新以恢复有效的策略。这种严重问题导致高样本低效率和对新情况的适应性差。为缓解此问题，我们提出了一种新的政策梯度估计，用于软MAX策略，该估计在批评中利用批评中的偏差和奖励信号中存在的噪声来逃避策略参数空间的饱和区域。我们对匪徒和古典MDP基准测试任务进行了分析和实验，表明我们的估算变得更加坚固，以便对政策饱和度更加强大。

最近基于进化的零级优化方法和基于策略梯度的一阶方法是解决加强学习（RL）问题的两个有希望的替代方案。前者的方法与任意政策一起工作，依赖状态依赖和时间扩展的探索，具有健壮性的属性，但遭受了较高的样本复杂性，而后者的方法更有效，但仅限于可区分的政策，并且学习的政策是不太强大。为了解决这些问题，我们提出了一种新颖的零级演员 - 批评算法（ZOAC），该算法将这两种方法统一为派对演员 - 批判性结构，以保留两者的优势。 ZOAC在参数空间，一阶策略评估（PEV）和零订单策略改进（PIM）的参数空间中进行了推出集合，每次迭代中都会进行推出。我们使用不同类型的策略在广泛的挑战连续控制基准上进行广泛评估我们的方法，其中ZOAC优于零阶和一阶基线算法。

Adequately assigning credit to actions for future outcomes based on their contributions is a long-standing open challenge in Reinforcement Learning. The assumptions of the most commonly used credit assignment method are disadvantageous in tasks where the effects of decisions are not immediately evident. Furthermore, this method can only evaluate actions that have been selected by the agent, making it highly inefficient. Still, no alternative methods have been widely adopted in the field. Hindsight Credit Assignment is a promising, but still unexplored candidate, which aims to solve the problems of both long-term and counterfactual credit assignment. In this thesis, we empirically investigate Hindsight Credit Assignment to identify its main benefits, and key points to improve. Then, we apply it to factored state representations, and in particular to state representations based on the causal structure of the environment. In this setting, we propose a variant of Hindsight Credit Assignment that effectively exploits a given causal structure. We show that our modification greatly decreases the workload of Hindsight Credit Assignment, making it more efficient and enabling it to outperform the baseline credit assignment method on various tasks. This opens the way to other methods based on given or learned causal structures.

Off-policy learning is more unstable compared to on-policy learning in reinforcement learning (RL). One reason for the instability of off-policy learning is a discrepancy between the target ($\pi$) and behavior (b) policy distributions. The discrepancy between $\pi$ and b distributions can be alleviated by employing a smooth variant of the importance sampling (IS), such as the relative importance sampling (RIS). RIS has parameter $\beta\in[0, 1]$ which controls smoothness. To cope with instability, we present the first relative importance sampling-off-policy actor-critic (RIS-Off-PAC) model-free algorithms in RL. In our method, the network yields a target policy (the actor), a value function (the critic) assessing the current policy ($\pi$) using samples drawn from behavior policy. We use action value generated from the behavior policy in reward function to train our algorithm rather than from the target policy. We also use deep neural networks to train both actor and critic. We evaluated our algorithm on a number of Open AI Gym benchmark problems and demonstrate better or comparable performance to several state-of-the-art RL baselines.

尽管政策梯度方法的普及日益越来越大，但它们尚未广泛用于样品稀缺应用，例如机器人。通过充分利用可用信息，可以提高样本效率。作为强化学习中的关键部件，奖励功能通常仔细设计以引导代理商。因此，奖励功能通常是已知的，允许访问不仅可以访问标量奖励信号，而且允许奖励梯度。为了从奖励梯度中受益，之前的作品需要了解环境动态，这很难获得。在这项工作中，我们开发\ Textit {奖励政策梯度}估计器，这是一种新的方法，可以在不学习模型的情况下整合奖励梯度。绕过模型动态允许我们的估算器实现更好的偏差差异，这导致更高的样本效率，如经验分析所示。我们的方法还提高了在不同的Mujoco控制任务上的近端策略优化的性能。

从现有数据中学习最佳行为是加强学习（RL）中最重要的问题之一。这被称为RL中的“非政策控制”，其中代理的目标是根据从给定策略（称为行为策略）获得的数据计算最佳策略。由于最佳策略可能与行为策略有很大不同，因此与“政体”设置相比，学习最佳行为非常困难，在学习中将利用来自策略更新的新数据。这项工作提出了一种非政策的天然参与者批评算法，该算法利用州行动分布校正来处理外部行为和样本效率的自然政策梯度。具有收敛保证的现有基于天然梯度的参与者批评算法需要固定功能，以近似策略和价值功能。这通常会导致许多RL应用中的次级学习。另一方面，我们提出的算法利用兼容功能，使人们能够使用任意神经网络近似策略和价值功能，并保证收敛到本地最佳策略。我们通过将其与基准RL任务上的香草梯度参与者 - 批评算法进行比较，说明了提出的非政策自然梯度算法的好处。

资产分配（或投资组合管理）是确定如何最佳将有限预算的资金分配给一系列金融工具/资产（例如股票）的任务。这项研究调查了使用无模型的深RL代理应用于投资组合管理的增强学习（RL）的性能。我们培训了几个RL代理商的现实股票价格，以学习如何执行资产分配。我们比较了这些RL剂与某些基线剂的性能。我们还比较了RL代理，以了解哪些类别的代理表现更好。从我们的分析中，RL代理可以执行投资组合管理的任务，因为它们的表现明显优于基线代理（随机分配和均匀分配）。四个RL代理（A2C，SAC，PPO和TRPO）总体上优于最佳基线MPT。这显示了RL代理商发现更有利可图的交易策略的能力。此外，基于价值和基于策略的RL代理之间没有显着的性能差异。演员批评者的表现比其他类型的药物更好。同样，在政策代理商方面的表现要好，因为它们在政策评估方面更好，样品效率在投资组合管理中并不是一个重大问题。这项研究表明，RL代理可以大大改善资产分配，因为它们的表现优于强基础。基于我们的分析，在政策上，参与者批评的RL药物显示出最大的希望。

基于我们先前关于绿色仿真辅助政策梯度（GS-PG）的研究，重点是基于轨迹的重复使用，在本文中，我们考虑了无限 - 马尔可夫马尔可夫决策过程，并创建了一种新的重要性采样策略梯度优化的方法来支持动态决策制造。现有的GS-PG方法旨在从完整的剧集或过程轨迹中学习，这将其适用性限制在低数据状态和灵活的在线过程控制中。为了克服这一限制，提出的方法可以选择性地重复使用最相关的部分轨迹，即，重用单元基于每步或每次派遣的历史观察。具体而言，我们创建了基于混合的可能性比率（MLR）策略梯度优化，该优化可以利用不同行为政策下产生的历史状态行动转变中的信息。提出的减少差异经验重播（VRER）方法可以智能地选择和重复使用最相关的过渡观察，改善策略梯度估计并加速最佳政策的学习。我们的实证研究表明，它可以改善优化融合并增强最先进的政策优化方法的性能，例如Actor-Critic方法和近端政策优化。

Many problems involve the use of models which learn probability distributions or incorporate randomness in some way. In such problems, because computing the true expected gradient may be intractable, a gradient estimator is used to update the model parameters. When the model parameters directly affect a probability distribution, the gradient estimator will involve score function terms. This paper studies baselines, a variance reduction technique for score functions. Motivated primarily by reinforcement learning, we derive for the first time an expression for the optimal state-dependent baseline, the baseline which results in a gradient estimator with minimum variance. Although we show that there exist examples where the optimal baseline may be arbitrarily better than a value function baseline, we find that the value function baseline usually performs similarly to an optimal baseline in terms of variance reduction. Moreover, the value function can also be used for bootstrapping estimators of the return, leading to additional variance reduction. Our results give new insight and justification for why value function baselines and the generalized advantage estimator (GAE) work well in practice.

In value-based reinforcement learning methods such as deep Q-learning, function approximation errors are known to lead to overestimated value estimates and suboptimal policies. We show that this problem persists in an actor-critic setting and propose novel mechanisms to minimize its effects on both the actor and the critic. Our algorithm builds on Double Q-learning, by taking the minimum value between a pair of critics to limit overestimation. We draw the connection between target networks and overestimation bias, and suggest delaying policy updates to reduce per-update error and further improve performance. We evaluate our method on the suite of OpenAI gym tasks, outperforming the state of the art in every environment tested.

由于数据量增加，金融业的快速变化已经彻底改变了数据处理和数据分析的技术，并带来了新的理论和计算挑战。与古典随机控制理论和解决财务决策问题的其他分析方法相比，解决模型假设的财务决策问题，强化学习（RL）的新发展能够充分利用具有更少模型假设的大量财务数据并改善复杂的金融环境中的决策。该调查纸目的旨在审查最近的资金途径的发展和使用RL方法。我们介绍了马尔可夫决策过程，这是许多常用的RL方法的设置。然后引入各种算法，重点介绍不需要任何模型假设的基于价值和基于策略的方法。连接是用神经网络进行的，以扩展框架以包含深的RL算法。我们的调查通过讨论了这些RL算法在金融中各种决策问题中的应用，包括最佳执行，投资组合优化，期权定价和对冲，市场制作，智能订单路由和Robo-Awaring。

强化学习的主要方法是根据预期的回报将信贷分配给行动。但是，我们表明回报可能取决于政策，这可能会导致价值估计的过度差异和减慢学习的速度。取而代之的是，我们证明了优势函数可以解释为因果效应，并与因果关系共享相似的属性。基于此洞察力，我们提出了直接优势估计（DAE），这是一种可以对优势函数进行建模并直接从政策数据进行估算的新方法，同时同时最大程度地减少了返回的方差而无需（操作 - ）值函数。我们还通过显示如何无缝整合到DAE中来将我们的方法与时间差异方法联系起来。所提出的方法易于实施，并且可以通过现代参与者批评的方法很容易适应。我们对三个离散控制域进行经验评估DAE，并表明它可以超过广义优势估计（GAE），这是优势估计的强大基线，当将大多数环境应用于策略优化时。

Deep reinforcement learning is poised to revolutionise the field of AI and represents a step towards building autonomous systems with a higher level understanding of the visual world. Currently, deep learning is enabling reinforcement learning to scale to problems that were previously intractable, such as learning to play video games directly from pixels. Deep reinforcement learning algorithms are also applied to robotics, allowing control policies for robots to be learned directly from camera inputs in the real world. In this survey, we begin with an introduction to the general field of reinforcement learning, then progress to the main streams of value-based and policybased methods. Our survey will cover central algorithms in deep reinforcement learning, including the deep Q-network, trust region policy optimisation, and asynchronous advantage actor-critic. In parallel, we highlight the unique advantages of deep neural networks, focusing on visual understanding via reinforcement learning. To conclude, we describe several current areas of research within the field.