We revisit the domain of off-policy policy optimization in RL from the perspective of coordinate ascent. One commonly-used approach is to leverage the off-policy policy gradient to optimize a surrogate objective -- the total discounted in expectation return of the target policy with respect to the state distribution of the behavior policy. However, this approach has been shown to suffer from the distribution mismatch issue, and therefore significant efforts are needed for correcting this mismatch either via state distribution correction or a counterfactual method. In this paper, we rethink off-policy learning via Coordinate Ascent Policy Optimization (CAPO), an off-policy actor-critic algorithm that decouples policy improvement from the state distribution of the behavior policy without using the policy gradient. This design obviates the need for distribution correction or importance sampling in the policy improvement step of off-policy policy gradient. We establish the global convergence of CAPO with general coordinate selection and then further quantify the convergence rates of several instances of CAPO with popular coordinate selection rules, including the cyclic and the randomized variants of CAPO. We then extend CAPO to neural policies for a more practical implementation. Through experiments, we demonstrate that CAPO provides a competitive approach to RL in practice.
translated by 谷歌翻译
政策优化是设计强化学习算法的基本原则,一个例子是具有剪切的替代物镜(PPO-CLIP)的近端政策优化算法(PPO-CLIP),由于其简单性和有效性,该算法已被普遍用于深度强化学习。尽管具有出色的经验表现,但PPO-CLIP尚未通过最新的理论证明是合理的。在本文中,我们在神经功能近似下建立了PPO-CLIP的第一个全局收敛速率。我们确定分析PPO-CLIP的基本挑战并用两个核心思想解决:(i)我们从铰链损失的角度重新解释了PPO-CLIP,这将政策改进与解决铰链损失和铰链损失和铰链损失和铰链分类问题的联系联系起来。提供PPO-CLIP目标的广义版。 (ii)基于上面的观点,我们提出了一个两步的策略改进方案,该方案通过熵镜下降和基于回归的策略更新方案从复杂的神经策略参数借助复杂的神经策略参数化来促进收敛分析。此外,我们的理论结果提供了剪辑机理对PPO-CLIP收敛的影响的首次表征。通过实验,我们从经验上验证了PPO-CLIP的重新解释,并在各种RL基准任务上具有各种分类器的广义目标。
translated by 谷歌翻译
在本文中,我们在表格设置中建立了违法演员批评算法的全球最优性和收敛速度,而不使用密度比来校正行为政策的状态分布与目标政策之间的差异。我们的工作超出了现有的工作原理,最佳的策略梯度方法中的现有工作中使用确切的策略渐变来更新策略参数时,我们使用近似和随机更新步骤。我们的更新步骤不是渐变更新,因为我们不使用密度比以纠正状态分布,这与从业者做得好。我们的更新是近似的,因为我们使用学习的评论家而不是真正的价值函数。我们的更新是随机的,因为在每个步骤中,更新仅为当前状态操作对完成。此外,我们在分析中删除了现有作品的几个限制性假设。我们的工作中的核心是基于其均匀收缩性能的时源性Markov链中的通用随机近似算法的有限样本分析。
translated by 谷歌翻译
The softmax policy gradient (PG) method, which performs gradient ascent under softmax policy parameterization, is arguably one of the de facto implementations of policy optimization in modern reinforcement learning. For $\gamma$-discounted infinite-horizon tabular Markov decision processes (MDPs), remarkable progress has recently been achieved towards establishing global convergence of softmax PG methods in finding a near-optimal policy. However, prior results fall short of delineating clear dependencies of convergence rates on salient parameters such as the cardinality of the state space $\mathcal{S}$ and the effective horizon $\frac{1}{1-\gamma}$, both of which could be excessively large. In this paper, we deliver a pessimistic message regarding the iteration complexity of softmax PG methods, despite assuming access to exact gradient computation. Specifically, we demonstrate that the softmax PG method with stepsize $\eta$ can take \[ \frac{1}{\eta} |\mathcal{S}|^{2^{\Omega\big(\frac{1}{1-\gamma}\big)}} ~\text{iterations} \] to converge, even in the presence of a benign policy initialization and an initial state distribution amenable to exploration (so that the distribution mismatch coefficient is not exceedingly large). This is accomplished by characterizing the algorithmic dynamics over a carefully-constructed MDP containing only three actions. Our exponential lower bound hints at the necessity of carefully adjusting update rules or enforcing proper regularization in accelerating PG methods.
translated by 谷歌翻译
Reinforcement learning (RL) problems can be challenging without well-shaped rewards. Prior work on provably efficient RL methods generally proposes to address this issue with dedicated exploration strategies. However, another way to tackle this challenge is to reformulate it as a multi-task RL problem, where the task space contains not only the challenging task of interest but also easier tasks that implicitly function as a curriculum. Such a reformulation opens up the possibility of running existing multi-task RL methods as a more efficient alternative to solving a single challenging task from scratch. In this work, we provide a theoretical framework that reformulates a single-task RL problem as a multi-task RL problem defined by a curriculum. Under mild regularity conditions on the curriculum, we show that sequentially solving each task in the multi-task RL problem is more computationally efficient than solving the original single-task problem, without any explicit exploration bonuses or other exploration strategies. We also show that our theoretical insights can be translated into an effective practical learning algorithm that can accelerate curriculum learning on simulated robotic tasks.
translated by 谷歌翻译
我们考虑一个不当的强化学习设置,在该设置中,为学习者提供了$ M $的基本控制器,以进行未知的马尔可夫决策过程,并希望最佳地结合它们,以生产一个可能胜过每个基本基础的控制器。这对于在不匹配或模拟环境中学习的跨控制器进行调整可能很有用,可以为给定的目标环境获得良好的控制器,而试验相对较少。在此方面,我们提出了两种算法:(1)一种基于政策梯度的方法; (2)可以根据可用信息在基于简单的参与者(AC)方案和天然参与者(NAC)方案之间切换的算法。两种算法都在给定控制器的一类不当混合物上运行。对于第一种情况,我们得出融合率保证,假设访问梯度甲骨文。对于基于AC的方法,我们提供了基本AC案例中的固定点的收敛速率保证,并在NAC情况下为全球最优值提供了保证。 (i)稳定卡特柱的标准控制理论基准的数值结果; (ii)一个受约束的排队任务表明,即使可以使用的基本策略不稳定,我们的不当政策优化算法也可以稳定系统。
translated by 谷歌翻译
为了在许多因素动态影响输出轨迹的复杂随机系统上学习,希望有效利用从以前迭代中收集的历史样本中的信息来加速策略优化。经典的经验重播使代理商可以通过重复使用历史观察来记住。但是,处理所有观察结果的统一重复使用策略均忽略了不同样本的相对重要性。为了克服这一限制,我们提出了一个基于一般差异的经验重播(VRER)框架,该框架可以选择性地重复使用最相关的样本以改善策略梯度估计。这种选择性机制可以自适应地对过去的样品增加重量,这些样本更可能由当前目标分布产生。我们的理论和实证研究表明,提议的VRER可以加速学习最佳政策,并增强最先进的政策优化方法的性能。
translated by 谷歌翻译
为了在许多因素动态影响输出轨迹的复杂随机系统上学习,希望有效利用从以前迭代中收集的历史样本中的信息来加速策略优化。经典的经验重播使代理商可以通过重复使用历史观察来记住。但是,处理所有观察结果的统一重复使用策略均忽略了不同样本的相对重要性。为了克服这一限制,我们提出了一个基于一般差异的经验重播(VRER)框架,该框架可以选择性地重复使用最相关的样本以改善策略梯度估计。这种选择性机制可以自适应地对过去的样品增加重量,这些样本更可能由当前目标分布产生。我们的理论和实证研究表明,提议的VRER可以加速学习最佳政策,并增强最先进的政策优化方法的性能。
translated by 谷歌翻译
We propose a new policy gradient method, named homotopic policy mirror descent (HPMD), for solving discounted, infinite horizon MDPs with finite state and action spaces. HPMD performs a mirror descent type policy update with an additional diminishing regularization term, and possesses several computational properties that seem to be new in the literature. We first establish the global linear convergence of HPMD instantiated with Kullback-Leibler divergence, for both the optimality gap, and a weighted distance to the set of optimal policies. Then local superlinear convergence is obtained for both quantities without any assumption. With local acceleration and diminishing regularization, we establish the first result among policy gradient methods on certifying and characterizing the limiting policy, by showing, with a non-asymptotic characterization, that the last-iterate policy converges to the unique optimal policy with the maximal entropy. We then extend all the aforementioned results to HPMD instantiated with a broad class of decomposable Bregman divergences, demonstrating the generality of the these computational properties. As a by product, we discover the finite-time exact convergence for some commonly used Bregman divergences, implying the continuing convergence of HPMD to the limiting policy even if the current policy is already optimal. Finally, we develop a stochastic version of HPMD and establish similar convergence properties. By exploiting the local acceleration, we show that for small optimality gap, a better than $\tilde{\mathcal{O}}(\left|\mathcal{S}\right| \left|\mathcal{A}\right| / \epsilon^2)$ sample complexity holds with high probability, when assuming a generative model for policy evaluation.
translated by 谷歌翻译
在这项工作中,我们研究了解决强化学习问题的基于政策的方法,其中采用了非政策性采样和线性函数近似进行政策评估,以及包括自然政策梯度(NPG)在内的各种政策更新规则,用于政策更新。为了在致命三合会的存在下解决政策评估子问题,我们提出了一个通用算法的多步型TD学习框架,具有广义的重要性抽样比率,其中包括两个特定的算法:$ \ lambda $ Q Q $ Q Q $ - 跟踪和双面$ Q $ - 跟踪。通用算法是单个时间尺度,具有可证明的有限样本保证,并克服了非政策学习中的高方差问题。至于策略更新,我们仅使用Bellman操作员的收缩属性和单调性属性提供通用分析,以在各种策略更新规则下建立几何融合。重要的是,通过将NPG视为实施政策迭代的近似方法,我们在不引入正则化的情况下建立了NPG的几何融合,并且不使用现有文献中的镜像下降类型的分析类型。将策略更新的几何融合与策略评估的有限样本分析相结合,我们首次建立了整​​体$ \ Mathcal {o}(\ Epsilon^{ - 2})$样本复杂性以找到最佳策略(最多达到函数近似误差)使用基于策略的方法和线性函数近似下的基于策略的方法。
translated by 谷歌翻译
我们研究了平均奖励马尔可夫决策过程(AMDP)的问题,并开发了具有强大理论保证的新型一阶方法,以进行政策评估和优化。由于缺乏勘探,现有的彻底评估方法遭受了次优融合率以及处理不足的随机策略(例如确定性政策)的失败。为了解决这些问题,我们开发了一种新颖的差异时间差异(VRTD)方法,具有随机策略的线性函数近似以及最佳收敛保证,以及一种探索性方差降低的时间差(EVRTD)方法,用于不充分的随机策略,可相当的融合保证。我们进一步建立了政策评估偏见的线性收敛速率,这对于改善策略优化的总体样本复杂性至关重要。另一方面,与对MDP的政策梯度方法的有限样本分析相比,对AMDP的策略梯度方法的现有研究主要集中在基础马尔可夫流程的限制性假设下(例如,参见Abbasi-e, Yadkori等人,2019年),他们通常缺乏整体样本复杂性的保证。为此,我们开发了随机策略镜下降(SPMD)的平均奖励变体(LAN,2022)。我们建立了第一个$ \ widetilde {\ Mathcal {o}}(\ epsilon^{ - 2})$样品复杂性,用于在生成模型(带有UNICHAIN假设)和Markovian Noise模型(使用Ergodicicic Modele(具有核能的模型)下,使用策略梯度方法求解AMDP假设)。该界限可以进一步改进到$ \ widetilde {\ Mathcal {o}}}(\ epsilon^{ - 1})$用于求解正则化AMDPS。我们的理论优势通过数值实验来证实。
translated by 谷歌翻译
在标准数据分析框架中,首先收集数据(全部一次),然后进行数据分析。此外,通常认为数据生成过程是外源性的。当数据分析师对数据的生成方式没有影响时,这种方法是自然的。但是,数字技术的进步使公司促进了从数据中学习并同时做出决策。随着这些决定生成新数据,数据分析师(业务经理或算法)也成为数据生成器。这种相互作用会产生一种新型的偏见 - 增强偏见 - 加剧了静态数据分析中的内生性问题。因果推理技术应该被纳入加强学习中以解决此类问题。
translated by 谷歌翻译
由于策略梯度定理导致的策略设置存在各种理论上 - 声音策略梯度算法,其为梯度提供了简化的形式。然而,由于存在多重目标和缺乏明确的脱助政策政策梯度定理,截止策略设置不太明确。在这项工作中,我们将这些目标统一到一个违规目标,并为此统一目标提供了政策梯度定理。推导涉及强调的权重和利息职能。我们显示多种策略来近似梯度,以识别权重(ACE)称为Actor评论家的算法。我们证明了以前(半梯度)脱离政策演员 - 评论家 - 特别是offpac和DPG - 收敛到错误的解决方案,而Ace找到最佳解决方案。我们还强调为什么这些半梯度方法仍然可以在实践中表现良好,表明ace中的方差策略。我们经验研究了两个经典控制环境的若干ACE变体和基于图像的环境,旨在说明每个梯度近似的权衡。我们发现,通过直接逼近强调权重,ACE在所有测试的所有设置中执行或优于offpac。
translated by 谷歌翻译
我们研究具有多个奖励价值函数的马尔可夫决策过程(MDP)的政策优化,应根据给定的标准共同优化,例如比例公平(平滑凹面标量),硬约束(约束MDP)和Max-Min Trade-离开。我们提出了一个改变锚定的正规自然政策梯度(ARNPG)框架,该框架可以系统地将良好表现的一阶方法中的思想纳入多目标MDP问题的策略优化算法的设计。从理论上讲,基于ARNPG框架的设计算法实现了$ \ tilde {o}(1/t)$全局收敛,并具有精确的梯度。从经验上讲,与某些现有的基于策略梯度的方法相比,ARNPG引导的算法在精确梯度和基于样本的场景中也表现出卓越的性能。
translated by 谷歌翻译
策略梯度方法适用于复杂的,不理解的,通过对参数化的策略进行随机梯度下降来控制问题。不幸的是,即使对于可以通过标准动态编程技术解决的简单控制问题,策略梯度算法也会面临非凸优化问题,并且被广泛理解为仅收敛到固定点。这项工作确定了结构属性 - 通过几个经典控制问题共享 - 确保策略梯度目标函数尽管是非凸面,但没有次优的固定点。当这些条件得到加强时,该目标满足了产生收敛速率的Polyak-lojasiewicz(梯度优势)条件。当其中一些条件放松时,我们还可以在任何固定点的最佳差距上提供界限。
translated by 谷歌翻译
Reinforcement learning is a framework for interactive decision-making with incentives sequentially revealed across time without a system dynamics model. Due to its scaling to continuous spaces, we focus on policy search where one iteratively improves a parameterized policy with stochastic policy gradient (PG) updates. In tabular Markov Decision Problems (MDPs), under persistent exploration and suitable parameterization, global optimality may be obtained. By contrast, in continuous space, the non-convexity poses a pathological challenge as evidenced by existing convergence results being mostly limited to stationarity or arbitrary local extrema. To close this gap, we step towards persistent exploration in continuous space through policy parameterizations defined by distributions of heavier tails defined by tail-index parameter alpha, which increases the likelihood of jumping in state space. Doing so invalidates smoothness conditions of the score function common to PG. Thus, we establish how the convergence rate to stationarity depends on the policy's tail index alpha, a Holder continuity parameter, integrability conditions, and an exploration tolerance parameter introduced here for the first time. Further, we characterize the dependence of the set of local maxima on the tail index through an exit and transition time analysis of a suitably defined Markov chain, identifying that policies associated with Levy Processes of a heavier tail converge to wider peaks. This phenomenon yields improved stability to perturbations in supervised learning, which we corroborate also manifests in improved performance of policy search, especially when myopic and farsighted incentives are misaligned.
translated by 谷歌翻译
熵正则化是增强学习(RL)的流行方法。尽管它具有许多优势,但它改变了原始马尔可夫决策过程(MDP)的RL目标。尽管已经提出了差异正则化来解决这个问题,但不能微不足道地应用于合作的多代理增强学习(MARL)。在本文中,我们研究了合作MAL中的差异正则化,并提出了一种新型的非政策合作MARL框架,差异性的多代理参与者 - 参与者(DMAC)。从理论上讲,我们得出了DMAC的更新规则,该规则自然存在,并保证了原始MDP和Divergence regullatized MDP的单调政策改进和收敛。我们还给出了原始MDP中融合策略和最佳策略之间的差异。 DMAC是一个灵活的框架,可以与许多现有的MARL算法结合使用。从经验上讲,我们在教学随机游戏和Starcraft Multi-Agent挑战中评估了DMAC,并表明DMAC显着提高了现有的MARL算法的性能。
translated by 谷歌翻译
政策梯度(PG)算法是备受期待的强化学习对现实世界控制任务(例如机器人技术)的最佳候选人之一。但是,每当必须在物理系统上执行学习过程本身或涉及任何形式的人类计算机相互作用时,这些方法的反复试验性质就会提出安全问题。在本文中,我们解决了一种特定的安全公式,其中目标和危险都以标量奖励信号进行编码,并且学习代理被限制为从不恶化其性能,以衡量为预期的奖励总和。通过从随机优化的角度研究仅行为者的政策梯度,我们为广泛的参数政策建立了改进保证,从而将现有结果推广到高斯政策上。这与策略梯度估计器的差异的新型上限一起,使我们能够识别出具有很高概率的单调改进的元参数计划。两个关键的元参数是参数更新的步长和梯度估计的批处理大小。通过对这些元参数的联合自适应选择,我们获得了具有单调改进保证的政策梯度算法。
translated by 谷歌翻译
在本文中,我们提出了一种新的策略梯度(PG)方法,即用于解决一类正规化加固学习(RL)问题的块策略镜下降(BPMD)方法(强烈) - convex正规化器。与带有批处理更新规则的传统PG方法(访问和更新每个状态的策略)相比,BPMD方法通过部分更新规则具有廉价的每卷计算,该规则在采样状态上执行策略更新。尽管问题的性质和部分更新规则具有非概念性质,但我们还是为多种采样方案提供了统一的分析,并表明BPMD可以实现快速的线性收敛到全局最优性。特别是,均匀的采样导致可比的最坏情况总计算复杂性与批处理PG方法。还确定了一种与上policy采样的必要条件。通过混合采样方案,我们进一步表明,BPMD具有潜在的实例依赖性加速度,从而改善了对状态空间的依赖性,因此优于批次PG方法。然后,我们通过利用从样品构建的随机一阶信息扩展到随机设置。使用生成模型,$ \ tilde {\ mathcal {o}}(\ left \ lvert \ lerver \ mathcal {s} \ right \ rvert \ rvert \ left \ lest \ lerver \ lovt \ mathcal {a} \ right \ right \ rvert \ rvert \ rvert /\ epsilon) $ \ tilde {\ mathcal {o}}(\ left \ lvert \ m athcal {s} \ right \ rvert \ rvert \ left \ lest \ lvert \ lerver \ mathcal {a} \ right \ right \ rvert /\ epsilon^2)强率强度(分别为非巧克力符号)正规化器,其中$ \ epsilon $表示目标准确性。据我们所知,这是第一次开发和分析了块坐标下降方法,以进行强化学习的策略优化,这为解决大规模RL问题提供了新的观点。
translated by 谷歌翻译
In this work we introduce reinforcement learning techniques for solving lexicographic multi-objective problems. These are problems that involve multiple reward signals, and where the goal is to learn a policy that maximises the first reward signal, and subject to this constraint also maximises the second reward signal, and so on. We present a family of both action-value and policy gradient algorithms that can be used to solve such problems, and prove that they converge to policies that are lexicographically optimal. We evaluate the scalability and performance of these algorithms empirically, demonstrating their practical applicability. As a more specific application, we show how our algorithms can be used to impose safety constraints on the behaviour of an agent, and compare their performance in this context with that of other constrained reinforcement learning algorithms.
translated by 谷歌翻译