Deep neural networks provide Reinforcement Learning (RL) powerful function approximators to address large-scale decision-making problems. However, these approximators introduce challenges due to the non-stationary nature of RL training. One source of the challenges in RL is that output predictions can churn, leading to uncontrolled changes after each batch update for states not included in the batch. Although such a churn phenomenon exists in each step of network training, how churn occurs and impacts RL remains under-explored. In this work, we start by characterizing churn in a view of Generalized Policy Iteration with function approximation, and we discover a chain effect of churn that leads to a cycle where the churns in value estimation and policy improvement compound and bias the learning dynamics throughout the iteration. Further, we concretize the study and focus on the learning issues caused by the chain effect in different settings, including greedy action deviation in value-based methods, trust region violation in proximal policy optimization, and dual bias of policy value in actor-critic methods. We then propose a method to reduce the chain effect across different settings, called Churn Approximated ReductIoN (CHAIN), which can be easily plugged into most existing DRL algorithms. Our experiments demonstrate the effectiveness of our method in both reducing churn and improving learning performance across online and offline, value-based and policy-based RL settings, as well as a scaling setting.
We identify and study the phenomenon of policy churn, that is, the rapid change of the greedy policy in value-based reinforcement learning. Policy churn operates at a surprisingly rapid pace, changing the greedy action in a large fraction of states within a handful of learning updates (in a typical deep RL set-up such as DQN on Atari). We characterise the phenomenon empirically, verifying that it is not limited to specific algorithm or environment properties. A number of ablations help whittle down the plausible explanations on why churn occurs to just a handful, all related to deep learning. …
Google Research（谷歌研究院）；Princeton（普林斯顿大学）；Purdue University（普渡大学）；Carnegie Mellon University（卡内基梅隆大学）
We study the sample complexity of reducing reinforcement learning to a sequence of empirical risk minimization problems over the policy space. Such reductions-based algorithms exhibit local convergence in the function space, as opposed to the parameter space for policy gradient algorithms, and thus are unaffected by the possibly non-linear or discontinuous parameterization of the policy class. We propose a variance-reduced variant of Conservative Policy Iteration that improves the sample complexity of producing a $\varepsilon$-functional local optimum from $O(\varepsilon^{-4})$ to $O(\varepsil…
In offline reinforcement learning, a policy learns to maximize cumulative rewards with a fixed collection of data. Towards conservative strategy, current methods choose to regularize the behavior policy or learn a lower bound of the value function. However, exorbitant conservation tends to impair the policy's generalization ability and degrade its performance, especially for the mixed datasets. In this paper, we propose the method of reducing conservativeness oriented reinforcement learning. On the one hand, the policy is trained to pay more attention to the minority samples in the static data…
This paper investigates the use of prior computation to estimate the value function to improve sample efficiency in on-policy policy gradient methods in reinforcement learning. Our approach is to estimate the value function from prior computations, such as from the Q-network learned in DQN or the value function trained for different but related environments. In particular, we learn a new value function for the target task while combining it with a value estimate from the prior computation. Finally, the resulting value function is used as a baseline in the policy gradient method. This use of a …
McCombs School of Business（麦库姆斯商学院）；The University of Texas at Austin（德克萨斯大学奥斯汀分校）
Offline reinforcement learning (RL) extends the paradigm of classical RL algorithms to purely learning from static datasets, without interacting with the underlying environment during the learning process. A key challenge of offline RL is the instability of policy training, caused by the mismatch between the distribution of the offline data and the undiscounted stationary state-action distribution of the learned policy. To avoid the detrimental impact of distribution mismatch, we regularize the undiscounted stationary distribution of the current policy towards the offline data during the polic…
RollVerify: Bridging Efficiency and Accuracy in Long-Tail Rollout Reinforcement Learning
Shanghai Jiao Tong University（上海交通大学）；Central South University（中南大学）；SenseTime Research（商汤科技研究院）；The Chinese University of Hong Kong（香港中文大学）；Beihang University（北京航空航天大学）
Reducing cross-sample prediction churn in scientific machine learning
Helmholtz-Zentrum Berlin für Materialien und Energie GmbH（柏林材料与能源研究中心）；Helmholtz Institute for Polymers in Energy Applications Jena（耶拿聚合物能源应用赫尔姆霍茨研究所）；Laboratory of Organic and Macromolecular Chemistry (IOMC)（有机与大分子化学实验室）；Center for Energy and Environmental Chemistry Jena（耶拿能源与环境化学中心）；Jena Center for Soft Matter (JCSM)（耶拿软物质研究中心）
University of Alberta（阿尔伯塔大学）；Alberta Machine Intelligence Institute (Amii)（阿尔伯塔机器智能研究所（Amii））
Approximate Policy Iteration (API) algorithms alternate between (approximate) policy evaluation and (approximate) greedification. Many different approaches have been explored for approximate policy evaluation, but less is understood about approximate greedification and what choices guarantee policy improvement. In this work, we investigate approximate greedification when reducing the KL divergence between the parameterized policy and the Boltzmann distribution over action values. In particular, we investigate the difference between the forward and reverse KL divergences, with varying degrees o…
Standard training techniques for neural networks involve multiple sources of randomness, e.g., initialization, mini-batch ordering and in some cases data augmentation. Given that neural networks are heavily over-parameterized in practice, such randomness can cause {\em churn} -- for the same input, disagreements between predictions of the two models independently trained by the same algorithm, contributing to the `reproducibility challenges' in modern machine learning. In this paper, we study this problem of churn, identify factors that cause it, and propose two simple means of mitigating it. …
The University of Texas at Austin（德克萨斯大学奥斯汀分校）；School of Informatics, University of Edinburgh（爱丁堡大学信息学院）；University of Wisconsin–Madison（威斯康星大学麦迪逊分校）；Sony AI（索尼人工智能）
Temporal difference (TD) learning is one of the main foundations of modern reinforcement learning. This paper studies the use of TD(0), a canonical TD algorithm, to estimate the value function of a given policy from a batch of data. In this batch setting, we show that TD(0) may converge to an inaccurate value function because the update following an action is weighted according to the number of times that action occurred in the batch -- not the true probability of the action under the given policy. To address this limitation, we introduce \textit{policy sampling error corrected}-TD(0) (PSEC-TD…
We present a simple, yet powerful data-augmentation technique to enable data-efficient learning from parametric experts for reinforcement and imitation learning. We focus on what we call the policy cloning setting, in which we use online or offline queries of an expert or expert policy to inform the behavior of a student policy. This setting arises naturally in a number of problems, for instance as variants of behavior cloning, or as a component of other algorithms such as DAGGER, policy distillation or KL-regularized RL. Our approach, augmented policy cloning (APC), uses synthetic states to i…
While often stated as an instance of the likelihood ratio trick [Rubinstein, 1989], the original policy gradient theorem [Sutton, 1999] involves an integral over the action space. When this integral can be computed, the resulting "all-action" estimator [Sutton, 2001] provides a conditioning effect [Bratley, 1987] reducing the variance significantly compared to the REINFORCE estimator [Williams, 1992]. In this paper, we adopt a numerical integration perspective to broaden the applicability of the all-action estimator to general spaces and to any function class for the policy or critic component…
On-policy reinforcement learning (RL) algorithms have high sample complexity while off-policy algorithms are difficult to tune. Merging the two holds the promise to develop efficient algorithms that generalize across diverse environments. It is however challenging in practice to find suitable hyper-parameters that govern this trade off. This paper develops a simple algorithm named P3O that interleaves off-policy updates with on-policy updates. P3O uses the effective sample size between the behavior policy and the target policy to control how far they can be from each other and does not introdu…
University of Illinois, Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）
In this paper, we present a new approach to Transfer Learning (TL) in Reinforcement Learning (RL) for cross-domain tasks. Many of the available techniques approach the transfer architecture as a method of speeding up the target task learning. We propose to adapt and reuse the mapped source task optimal-policy directly in related domains. We show the optimal policy from a related source task can be near optimal in target domain provided an adaptive policy accounts for the model error between target and source. The main benefit of this policy augmentation is generalizing policies across multiple…
A central challenge to applying many off-policy reinforcement learning algorithms to real world problems is the variance introduced by importance sampling. In off-policy learning, the agent learns about a different policy than the one being executed. To account for the difference importance sampling ratios are often used, but can increase variance in the algorithms and reduce the rate of learning. Several variations of importance sampling have been proposed to reduce variance, with per-decision importance sampling being the most popular. However, the update rules for most off-policy algorithms…
Off-policy evaluation is a key component of reinforcement learning which evaluates a target policy with offline data collected from behavior policies. It is a crucial step towards safe reinforcement learning and has been used in advertisement, recommender systems and many other applications. In these applications, sometimes the offline data is collected from multiple behavior policies. Previous works regard data from different behavior policies equally. Nevertheless, some behavior policies are better at producing good estimators while others are not. This paper starts with discussing how to co…
Robotics Design Laboratory, University of Queensland（昆士兰大学机器人设计实验室）
Soft Actor-Critic (SAC) is one of the state-of-the-art off-policy reinforcement learning (RL) algorithms that is within the maximum entropy based RL framework. SAC is demonstrated to perform very well in a list of continous control tasks with good stability and robustness. SAC learns a stochastic Gaussian policy that can maximize a trade-off between total expected reward and the policy entropy. To update the policy, SAC minimizes the KL-Divergence between the current policy density and the soft value function density. Reparameterization trick is then used to obtain the approximate gradient of …
In reinforcement learning (RL), function approximation errors are known to easily lead to the Q-value overestimations, thus greatly reducing policy performance. This paper presents a distributional soft actor-critic (DSAC) algorithm, which is an off-policy RL method for continuous control setting, to improve the policy performance by mitigating Q-value overestimations. We first discover in theory that learning a distribution function of state-action returns can effectively mitigate Q-value overestimations because it is capable of adaptively adjusting the update stepsize of the Q-value function…
We propose a successive convex approximation based off-policy optimization (SCAOPO) algorithm to solve the general constrained reinforcement learning problem, which is formulated as a constrained Markov decision process (CMDP) in the context of average cost. The SCAOPO is based on solving a sequence of convex objective/feasibility optimization problems obtained by replacing the objective and constraint functions in the original problems with convex surrogate functions. At each iteration, the convex surrogate problem can be efficiently solved by Lagrange dual method even the policy is parameter…
Warwick Medical School, University of Warwick（华威大学华威医学院）；WMG, University of Warwick（华威大学华威制造集团）；Department of Statistics, University of Warwick（华威大学统计系）；Alan Turing Institute（艾伦·图灵研究所）
Offline reinforcement learning agents seek optimal policies from fixed data sets. With environmental interaction prohibited, agents face significant challenges in preventing errors in value estimates from compounding and subsequently causing the learning process to collapse. Uncertainty estimation using ensembles compensates for this by penalising high-variance value estimates, allowing agents to learn robust policies based on data-driven actions. However, the requirement for large ensembles to facilitate sufficient penalisation results in significant computational overhead. In this work, we e…
Solving the Goal-Conditioned Reward Sparse (GCRS) task is a challenging reinforcement learning problem due to the sparsity of reward signals. In this work, we propose a new formulation of GCRS tasks from the perspective of the drifted random walk on the state space, and design a novel method called Evolutionary Stochastic Policy Distillation (ESPD) to solve them based on the insight of reducing the First Hitting Time of the stochastic process. As a self-imitate approach, ESPD enables a target policy to learn from a series of its stochastic variants through the technique of policy distillation …
Recent developments have established the vulnerability of deep reinforcement learning to policy manipulation attacks via intentionally perturbed inputs, known as adversarial examples. In this work, we propose a technique for mitigation of such attacks based on addition of noise to the parameter space of deep reinforcement learners during training. We experimentally verify the effect of parameter-space noise in reducing the transferability of adversarial examples, and demonstrate the promising performance of this technique in mitigating the impact of whitebox and blackbox attacks at both test a…
The bus system is a critical component of sustainable urban transportation. However, due to the significant uncertainties in passenger demand and traffic conditions, bus operation is unstable in nature and bus bunching has become a common phenomenon that undermines the reliability and efficiency of bus services. Despite recent advances in multi-agent reinforcement learning (MARL) on traffic control, little research has focused on bus fleet control due to the tricky asynchronous characteristic -- control actions only happen when a bus arrives at a bus stop and thus agents do not act simultaneou…
Deep reinforcement learning (DRL) has been demonstrated to provide promising results in several challenging decision making and control tasks. However, the required inference costs of deep neural networks (DNNs) could prevent DRL from being applied to mobile robots which cannot afford high energy-consuming computations. To enable DRL methods to be affordable in such energy-limited platforms, we propose an asymmetric architecture that reduces the overall inference costs via switching between a computationally expensive policy and an economic one. The experimental results evaluated on a number o…
