Offline reinforcement learning (RL), which refers to decision-making from a previously-collected dataset of interactions, has received significant attention over the past years. Much effort has focused on improving offline RL practicality by addressing the prevalent issue of partial data coverage through various forms of conservative policy learning. While the majority of algorithms do not have finite-sample guarantees, several provable conservative offline RL algorithms are designed and analyzed within the single-policy concentrability framework that handles partial coverage. Yet, in the nonlinear function approximation setting where confidence intervals are difficult to obtain, existing provable algorithms suffer from computational intractability, prohibitively strong assumptions, and suboptimal statistical rates. In this paper, we leverage the marginalized importance sampling (MIS) formulation of RL and present the first set of offline RL algorithms that are statistically optimal and practical under general function approximation and single-policy concentrability, bypassing the need for uncertainty quantification. We identify that the key to successfully solving the sample-based approximation of the MIS problem is ensuring that certain occupancy validity constraints are nearly satisfied. We enforce these constraints by a novel application of the augmented Lagrangian method and prove the following result: with the MIS formulation, augmented Lagrangian is enough for statistically optimal offline RL. In stark contrast to prior algorithms that induce additional conservatism through methods such as behavior regularization, our approach provably eliminates this need and reinterprets regularizers as "enforcers of occupancy validity" than "promoters of conservatism."
相關內容
This project leverages advances in multi-agent reinforcement learning (MARL) to improve the efficiency and flexibility of order-picking systems for commercial warehouses. We envision a warehouse of the future in which dozens of mobile robots and human pickers work together to collect and deliver items within the warehouse. The fundamental problem we tackle, called the order-picking problem, is how these worker agents must coordinate their movement and actions in the warehouse to maximise performance (e.g. order throughput) under given resource constraints. Established industry methods using heuristic approaches require large engineering efforts to optimise for innately variable warehouse configurations. In contrast, the MARL framework can be flexibly applied to any warehouse configuration (e.g. size, layout, number/types of workers, item replenishment frequency) and the agents learn via a process of trial-and-error how to optimally cooperate with one another. This paper details the current status of the R&D effort initiated by Dematic and the University of Edinburgh towards a general-purpose and scalable MARL solution for the order-picking problem in realistic warehouses.
Recommender systems aim to answer the following question: given the items that a user has interacted with, what items will this user likely interact with next? Historically this problem is often framed as a predictive task via (self-)supervised learning. In recent years, we have seen more emphasis placed on approaching the recommendation problem from a policy optimization perspective: learning a policy that maximizes some reward function (e.g., user engagement). However, it is commonly the case in recommender systems that we are only able to train a new policy given data collected from a previously-deployed policy. The conventional way to address such a policy mismatch is through importance sampling correction, which unfortunately comes with its own limitations. In this paper, we suggest an alternative approach, which involves the use of local policy improvement without off-policy correction. Drawing from a number of related results in the fields of causal inference, bandits, and reinforcement learning, we present a suite of methods that compute and optimize a lower bound of the expected reward of the target policy. Crucially, this lower bound is a function that is easy to estimate from data, and which does not involve density ratios (such as those appearing in importance sampling correction). We argue that this local policy improvement paradigm is particularly well suited for recommender systems, given that in practice the previously-deployed policy is typically of reasonably high quality, and furthermore it tends to be re-trained frequently and gets continuously updated. We discuss some practical recipes on how to apply some of the proposed techniques in a sequential recommendation setting.
In off-policy reinforcement learning, a behaviour policy performs exploratory interactions with the environment to obtain state-action-reward samples which are then used to learn a target policy that optimises the expected return. This leads to a problem of off-policy evaluation, where one needs to evaluate the target policy from samples collected by the often unrelated behaviour policy. Importance sampling is a traditional statistical technique that is often applied to off-policy evaluation. While importance sampling estimators are unbiased, their variance increases exponentially with the horizon of the decision process due to computing the importance weight as a product of action probability ratios, yielding estimates with low accuracy for domains involving long-term planning. This paper proposes state-based importance sampling (SIS), which drops the action probability ratios of sub-trajectories with "neglible states" -- roughly speaking, those for which the chosen actions have no impact on the return estimate -- from the computation of the importance weight. Theoretical results show that this results in a reduction of the exponent in the variance upper bound as well as improving the mean squared error. An automated search algorithm based on covariance testing is proposed to identify a negligible state set which has minimal MSE when performing state-based importance sampling. Experiments are conducted on a lift domain, which include "lift states" where the action has no impact on the following state and reward. The results demonstrate that using the search algorithm, SIS yields reduced variance and improved accuracy compared to traditional importance sampling, per-decision importance sampling, and incremental importance sampling.
Research questions across a diverse array of fields are formulated as a Partial Conjunction Hypothesis (PCH) tests, which combine information across $m$ base hypotheses to determine whether some subset is non-null. However, standard methods for testing a PCH can be highly conservative. In this paper, we introduce the conditional PCH (cPCH) test, a new framework for testing a single PCH that directly corrects the conservativeness of standard approaches by conditioning on certain order statistics of the base p-values. Under distributional assumptions commonly encountered in PCH testing, the cPCH test produces a p-value that is very nearly uniform. Through simulations, we demonstrate that the cPCH test uniformly outperforms standard single PCH tests and maintains Type I error control even under model misspecification, and can in certain situations also be used to outperform state-of-the-art PCH multiple testing procedures. Finally, we illustrate an application of the cPCH test on a replicability analysis of four microarray studies.
Reinforcement Learning is a powerful tool to model decision-making processes. However, it relies on an exploration-exploitation trade-off that remains an open challenge for many tasks. In this work, we study neighboring state-based, model-free exploration led by the intuition that, for an early-stage agent, considering actions derived from a bounded region of nearby states may lead to better actions when exploring. We propose two algorithms that choose exploratory actions based on a survey of nearby states, and find that one of our methods, ${\rho}$-explore, consistently outperforms the Double DQN baseline in an discrete environment by 49\% in terms of Eval Reward Return.
Task transfer learning is a popular technique in image processing applications that uses pre-trained models to reduce the supervision cost of related tasks. An important question is to determine task transferability, i.e. given a common input domain, estimating to what extent representations learned from a source task can help in learning a target task. Typically, transferability is either measured experimentally or inferred through task relatedness, which is often defined without a clear operational meaning. In this paper, we present a novel metric, H-score, an easily-computable evaluation function that estimates the performance of transferred representations from one task to another in classification problems using statistical and information theoretic principles. Experiments on real image data show that our metric is not only consistent with the empirical transferability measurement, but also useful to practitioners in applications such as source model selection and task transfer curriculum learning.
Any strategy used to distribute a robot ensemble over a set of sequential tasks is subject to inaccuracy due to robot-level uncertainties and environmental influences on the robots' behavior. We approach the problem of inaccuracy during task allocation by modeling and controlling the overall ensemble behavior. Our model represents the allocation problem as a stochastic jump process and we regulate the mean and variance of such a process. The main contributions of this paper are: Establishing a structure for the transition rates of the equivalent stochastic jump process and formally showing that this approach leads to decoupled parameters that allow us to adjust the first- and second-order moments of the ensemble distribution over tasks, which gives the flexibility to decrease the variance in the desired final distribution. This allows us to directly shape the impact of uncertainties on the group allocation over tasks. We introduce a detailed procedure to design the gains to achieve the desired mean and show how the additional parameters impact the covariance matrix, which is directly associated with the degree of task allocation precision. Our simulation and experimental results illustrate the successful control of several robot ensembles during task allocation.
In large-scale systems there are fundamental challenges when centralised techniques are used for task allocation. The number of interactions is limited by resource constraints such as on computation, storage, and network communication. We can increase scalability by implementing the system as a distributed task-allocation system, sharing tasks across many agents. However, this also increases the resource cost of communications and synchronisation, and is difficult to scale. In this paper we present four algorithms to solve these problems. The combination of these algorithms enable each agent to improve their task allocation strategy through reinforcement learning, while changing how much they explore the system in response to how optimal they believe their current strategy is, given their past experience. We focus on distributed agent systems where the agents' behaviours are constrained by resource usage limits, limiting agents to local rather than system-wide knowledge. We evaluate these algorithms in a simulated environment where agents are given a task composed of multiple subtasks that must be allocated to other agents with differing capabilities, to then carry out those tasks. We also simulate real-life system effects such as networking instability. Our solution is shown to solve the task allocation problem to 6.7% of the theoretical optimal within the system configurations considered. It provides 5x better performance recovery over no-knowledge retention approaches when system connectivity is impacted, and is tested against systems up to 100 agents with less than a 9% impact on the algorithms' performance.
Deep reinforcement learning algorithms can perform poorly in real-world tasks due to the discrepancy between source and target environments. This discrepancy is commonly viewed as the disturbance in transition dynamics. Many existing algorithms learn robust policies by modeling the disturbance and applying it to source environments during training, which usually requires prior knowledge about the disturbance and control of simulators. However, these algorithms can fail in scenarios where the disturbance from target environments is unknown or is intractable to model in simulators. To tackle this problem, we propose a novel model-free actor-critic algorithm -- namely, state-conservative policy optimization (SCPO) -- to learn robust policies without modeling the disturbance in advance. Specifically, SCPO reduces the disturbance in transition dynamics to that in state space and then approximates it by a simple gradient-based regularizer. The appealing features of SCPO include that it is simple to implement and does not require additional knowledge about the disturbance or specially designed simulators. Experiments in several robot control tasks demonstrate that SCPO learns robust policies against the disturbance in transition dynamics.
Bid optimization for online advertising from single advertiser's perspective has been thoroughly investigated in both academic research and industrial practice. However, existing work typically assume competitors do not change their bids, i.e., the wining price is fixed, leading to poor performance of the derived solution. Although a few studies use multi-agent reinforcement learning to set up a cooperative game, they still suffer the following drawbacks: (1) They fail to avoid collusion solutions where all the advertisers involved in an auction collude to bid an extremely low price on purpose. (2) Previous works cannot well handle the underlying complex bidding environment, leading to poor model convergence. This problem could be amplified when handling multiple objectives of advertisers which are practical demands but not considered by previous work. In this paper, we propose a novel multi-objective cooperative bid optimization formulation called Multi-Agent Cooperative bidding Games (MACG). MACG sets up a carefully designed multi-objective optimization framework where different objectives of advertisers are incorporated. A global objective to maximize the overall profit of all advertisements is added in order to encourage better cooperation and also to protect self-bidding advertisers. To avoid collusion, we also introduce an extra platform revenue constraint. We analyze the optimal functional form of the bidding formula theoretically and design a policy network accordingly to generate auction-level bids. Then we design an efficient multi-agent evolutionary strategy for model optimization. Offline experiments and online A/B tests conducted on the Taobao platform indicate both single advertiser's objective and global profit have been significantly improved compared to state-of-art methods.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191