In this paper, we present XuanCe, a comprehensive and unified deep reinforcement learning (DRL) library designed to be compatible with PyTorch, TensorFlow, and MindSpore. XuanCe offers a wide range of functionalities, including over 40 classical DRL and multi-agent DRL algorithms, with the flexibility to easily incorporate new algorithms and environments. It is a versatile DRL library that supports CPU, GPU, and Ascend, and can be executed on various operating systems such as Ubuntu, Windows, MacOS, and EulerOS. Extensive benchmarks conducted on popular environments including MuJoCo, Atari, and StarCraftII multi-agent challenge demonstrate the library's impressive performance. XuanCe is open-source and can be accessed at https://github.com/agi-brain/xuance.git.
Large-scale multiagent reinforcement learning requires huge computation and space costs, and the too-long execution process makes it hard to train policies for agents. This work proposes a concept of fuzzy agent, which is a new paradigm for training homogeneous agents. Aiming at a lightweight and affordable reinforcement learning mechanism for large-scale homogeneous multiagent systems, we break the one-to-one correspondence between agent and policy, designing abstract agents as the substitute for the multiagent to interact with the environment. The Markov decision process models for these abstract agents are conducted by fuzzy logic, which also acts on the behavior mapping from abstract agent to entity. Specifically, just the abstract agents execute their policy at a time step, and the concrete behaviors are generated by simple matrix operations. The proposal has lower space and computation complexities because the number of abstract agents is far less than that of entities, and the coupling among agents is retained implicitly. Compared with other approximation and simplification methods, the proposed fuzzy agent not only greatly reduces required computing resources but also ensures the effectiveness of the learned policies. Several experiments are conducted to validate our method. The results show that the proposal outperforms the baseline methods, while it has satisfactory zero-shot and few-shot transfer abilities.
Artificial Intelligence (AI) and Machine Learning (ML) are transforming the domain of Quantitative Trading (QT) through the deployment of advanced algorithms capable of sifting through extensive financial datasets to pinpoint lucrative investment openings. AI-driven models, particularly those employing ML techniques such as deep learning and reinforcement learning, have shown great prowess in predicting market trends and executing trades at a speed and accuracy that far surpass human capabilities. Its capacity to automate critical tasks, such as discerning market conditions and executing trading strategies, has been pivotal. However, persistent challenges exist in current QT methods, especially in effectively handling noisy and high-frequency financial data. Striking a balance between exploration and exploitation poses another challenge for AI-driven trading agents. To surmount these hurdles, our proposed solution, QTNet, introduces an adaptive trading model that autonomously formulates QT strategies through an intelligent trading agent. Incorporating deep reinforcement learning (DRL) with imitative learning methodologies, we bolster the proficiency of our model. To tackle the challenges posed by volatile financial datasets, we conceptualize the QT mechanism within the framework of a Partially Observable Markov Decision Process (POMDP). Moreover, by embedding imitative learning, the model can capitalize on traditional trading tactics, nurturing a balanced synergy between discovery and utilization. For a more realistic simulation, our trading agent undergoes training using minute-frequency data sourced from the live financial market. Experimental findings underscore the model's proficiency in extracting robust market features and its adaptability to diverse market conditions.
Efficient allocation is important in nature and human society, where individuals frequently compete for limited resources. The Minority Game (MG) is perhaps the simplest toy model to address this issue. However, most previous solutions assume that the strategies are provided a priori and static, failing to capture their adaptive nature. Here, we introduce the reinforcement learning (RL) paradigm to MG, where individuals adjust decisions based on accumulated experience and expected rewards dynamically. We find that this RL framework achieves optimal resource coordination when individuals balance the exploitation of experience with random exploration. Yet, the imbalanced strategies of the two lead to suboptimal partial coordination or even anti-coordination. Our mechanistic analysis reveals a symmetry-breaking in action preferences at the optimum, offering a fresh solution to the MG and new insights into the resource allocation problem.
In this paper, we consider federated reinforcement learning for tabular episodic Markov Decision Processes (MDP) where, under the coordination of a central server, multiple agents collaboratively explore the environment and learn an optimal policy without sharing their raw data. While linear speedup in the number of agents has been achieved for some metrics, such as convergence rate and sample complexity, in similar settings, it is unclear whether it is possible to design a model-free algorithm to achieve linear regret speedup with low communication cost. We propose two federated Q-Learning algorithms termed as FedQ-Hoeffding and FedQ-Bernstein, respectively, and show that the corresponding total regrets achieve a linear speedup compared with their single-agent counterparts when the time horizon is sufficiently large, while the communication cost scales logarithmically in the total number of time steps $T$. Those results rely on an event-triggered synchronization mechanism between the agents and the server, a novel step size selection when the server aggregates the local estimates of the state-action values to form the global estimates, and a set of new concentration inequalities to bound the sum of non-martingale differences. This is the first work showing that linear regret speedup and logarithmic communication cost can be achieved by model-free algorithms in federated reinforcement learning.
Effective communication protocols in multi-agent reinforcement learning (MARL) are critical to fostering cooperation and enhancing team performance. To leverage communication, many previous works have proposed to compress local information into a single message and broadcast it to all reachable agents. This simplistic messaging mechanism, however, may fail to provide adequate, critical, and relevant information to individual agents, especially in severely bandwidth-limited scenarios. This motivates us to develop context-aware communication schemes for MARL, aiming to deliver personalized messages to different agents. Our communication protocol, named CACOM, consists of two stages. In the first stage, agents exchange coarse representations in a broadcast fashion, providing context for the second stage. Following this, agents utilize attention mechanisms in the second stage to selectively generate messages personalized for the receivers. Furthermore, we employ the learned step size quantization (LSQ) technique for message quantization to reduce the communication overhead. To evaluate the effectiveness of CACOM, we integrate it with both actor-critic and value-based MARL algorithms. Empirical results on cooperative benchmark tasks demonstrate that CACOM provides evident performance gains over baselines under communication-constrained scenarios. The code is publicly available at https://github.com/LXXXXR/CACOM.
This paper studies a discrete-time mean-variance model based on reinforcement learning. Compared with its continuous-time counterpart in \cite{zhou2020mv}, the discrete-time model makes more general assumptions about the asset's return distribution. Using entropy to measure the cost of exploration, we derive the optimal investment strategy, whose density function is also Gaussian type. Additionally, we design the corresponding reinforcement learning algorithm. Both simulation experiments and empirical analysis indicate that our discrete-time model exhibits better applicability when analyzing real-world data than the continuous-time model.
Online safe reinforcement learning (RL) involves training a policy that maximizes task efficiency while satisfying constraints via interacting with the environments. In this paper, our focus lies in addressing the complex challenges associated with solving multi-constraint (MC) safe RL problems. We approach the safe RL problem from the perspective of Multi-Objective Optimization (MOO) and propose a unified framework designed for MC safe RL algorithms. This framework highlights the manipulation of gradients derived from constraints. Leveraging insights from this framework and recognizing the significance of \textit{redundant} and \textit{conflicting} constraint conditions, we introduce the Gradient Shaping (GradS) method for general Lagrangian-based safe RL algorithms to improve the training efficiency in terms of both reward and constraint satisfaction. Our extensive experimentation demonstrates the effectiveness of our proposed method in encouraging exploration and learning a policy that improves both safety and reward performance across various challenging MC safe RL tasks as well as good scalability to the number of constraints.
In this paper, we study asynchronous stochastic approximation algorithms without communication delays. Our main contribution is a stability proof for these algorithms that extends a method of Borkar and Meyn by accommodating more general noise conditions. We also derive convergence results from this stability result and discuss their application in important average-reward reinforcement learning problems.
Reinforcement Learning (RL), recognized as an efficient learning approach, has achieved remarkable success across multiple fields and applications, including gaming, robotics, and autonomous vehicles. Classical single-agent reinforcement learning grapples with the imbalance of exploration and exploitation as well as limited generalization abilities. This methodology frequently leads to algorithms settling for suboptimal solutions that are tailored only to specific datasets. In this work, our aim is to train agent with efficient learning by decoupling exploration and utilization, so that agent can escaping the conundrum of suboptimal Solutions. In reinforcement learning, the previously imposed pessimistic punitive measures have deprived the model of its exploratory potential, resulting in diminished exploration capabilities. To address this, we have introduced an additional optimistic Actor to enhance the model's exploration ability, while employing a more constrained pessimistic Actor for performance evaluation. The above idea is implemented in the proposed OPARL (Optimistic and Pessimistic Actor Reinforcement Learning) algorithm. This unique amalgamation within the reinforcement learning paradigm fosters a more balanced and efficient approach. It facilitates the optimization of policies that concentrate on high-reward actions via pessimistic exploitation strategies while concurrently ensuring extensive state coverage through optimistic exploration. Empirical and theoretical investigations demonstrate that OPARL enhances agent capabilities in both utilization and exploration. In the most tasks of DMControl benchmark and Mujoco environment, OPARL performed better than state-of-the-art methods. Our code has released on https://github.com/yydsok/OPARL
Machine unlearning refers to the process of mitigating the influence of specific training data on machine learning models based on removal requests from data owners. However, one important area that has been largely overlooked in the research of unlearning is reinforcement learning. Reinforcement learning focuses on training an agent to make optimal decisions within an environment to maximize its cumulative rewards. During the training, the agent tends to memorize the features of the environment, which raises a significant concern about privacy. As per data protection regulations, the owner of the environment holds the right to revoke access to the agent's training data, thus necessitating the development of a novel and pressing research field, known as \emph{reinforcement unlearning}. Reinforcement unlearning focuses on revoking entire environments rather than individual data samples. This unique characteristic presents three distinct challenges: 1) how to propose unlearning schemes for environments; 2) how to avoid degrading the agent's performance in remaining environments; and 3) how to evaluate the effectiveness of unlearning. To tackle these challenges, we propose two reinforcement unlearning methods. The first method is based on decremental reinforcement learning, which aims to erase the agent's previously acquired knowledge gradually. The second method leverages environment poisoning attacks, which encourage the agent to learn new, albeit incorrect, knowledge to remove the unlearning environment. Particularly, to tackle the third challenge, we introduce the concept of ``environment inference attack'' to evaluate the unlearning outcomes.
This research presents a novel active detection model utilizing deep reinforcement learning to accurately detect traffic objects in real-world scenarios. The model employs a deep Q-network based on LSTM-CNN that identifies and aligns target zones with specific categories of traffic objects through implementing a top-down approach with efficient feature extraction of the environment. The model integrates historical and current actions and observations to make a comprehensive analysis. The design of the state space and reward function takes into account the impact of time steps to enable the model to complete the task in fewer steps. Tests conducted demonstrate the model's proficiency, exhibiting exceptional precision and performance in locating traffic signal lights and speed limit signs. The findings of this study highlight the efficacy and potential of the deep reinforcement learning-based active detection model in traffic-related applications, underscoring its robust detection abilities and promising performance.
Problem definition: Supply chains are constantly evolving networks. Reinforcement learning is increasingly proposed as a solution to provide optimal control of these networks. Academic/practical: However, learning in continuously varying environments remains a challenge in the reinforcement learning literature.Methodology: This paper therefore seeks to address whether agents can control varying supply chain problems, transferring learning between environments that require different strategies and avoiding catastrophic forgetting of tasks that have not been seen in a while. To evaluate this approach, two state-of-the-art Reinforcement Learning (RL) algorithms are compared: an actor-critic learner, Proximal Policy Optimisation(PPO), and a Recurrent Proximal Policy Optimisation (RPPO), PPO with a Long Short-Term Memory(LSTM) layer, which is showing popularity in online learning environments. Results: First these methods are compared on six sets of environments with varying degrees of stochasticity. The results show that more lean strategies adopted in Batch environments are different from those adopted in Stochastic environments with varying products. The methods are also compared on various continuous supply chain scenarios, where the PPO agents are shown to be able to adapt through continuous learning when the tasks are similar but show more volatile performance when changing between the extreme tasks. However, the RPPO, with an ability to remember histories, is able to overcome this to some extent and takes on a more realistic strategy. Managerial implications: Our results provide a new perspective on the continuously varying supply chain, the cooperation and coordination of agents are crucial for improving the overall performance in uncertain and semi-continuous non-stationary supply chain environments without the need to retrain the environment as the demand changes.
Reinforcement learning has been demonstrated to outperform even the best humans in complex domains like video games. However, running reinforcement learning experiments on the required scale for autonomous driving is extremely difficult. Building a large scale reinforcement learning system and distributing it across many GPUs is challenging. Gathering experience during training on real world vehicles is prohibitive from a safety and scalability perspective. Therefore, an efficient and realistic driving simulator is required that uses a large amount of data from real-world driving. We bring these capabilities together and conduct large-scale reinforcement learning experiments for autonomous driving. We demonstrate that our policy performance improves with increasing scale. Our best performing policy reduces the failure rate by 64% while improving the rate of driving progress by 25% compared to the policies produced by state-of-the-art machine learning for autonomous driving.
Generalization and sample efficiency have been long-standing issues concerning reinforcement learning, and thus the field of Offline Meta-Reinforcement Learning~(OMRL) has gained increasing attention due to its potential of solving a wide range of problems with static and limited offline data. Existing OMRL methods often assume sufficient training tasks and data coverage to apply contrastive learning to extract task representations. However, such assumptions are not applicable in several real-world applications and thus undermine the generalization ability of the representations. In this paper, we consider OMRL with two types of data limitations: limited training tasks and limited behavior diversity and propose a novel algorithm called GENTLE for learning generalizable task representations in the face of data limitations. GENTLE employs Task Auto-Encoder~(TAE), which is an encoder-decoder architecture to extract the characteristics of the tasks. Unlike existing methods, TAE is optimized solely by reconstruction of the state transition and reward, which captures the generative structure of the task models and produces generalizable representations when training tasks are limited. To alleviate the effect of limited behavior diversity, we consistently construct pseudo-transitions to align the data distribution used to train TAE with the data distribution encountered during testing. Empirically, GENTLE significantly outperforms existing OMRL methods on both in-distribution tasks and out-of-distribution tasks across both the given-context protocol and the one-shot protocol.
Facility location problems on graphs are ubiquitous in real world and hold significant importance, yet their resolution is often impeded by NP-hardness. Recently, machine learning methods have been proposed to tackle such classical problems, but they are limited to the myopic constructive pattern and only consider the problems in Euclidean space. To overcome these limitations, we propose a general swap-based framework that addresses the p-median problem and the facility relocation problem on graphs and a novel reinforcement learning model demonstrating a keen awareness of complex graph structures. Striking a harmonious balance between solution quality and running time, our method surpasses handcrafted heuristics on intricate graph datasets. Additionally, we introduce a graph generation process to simulate real-world urban road networks with demand, facilitating the construction of large datasets for the classic problem. For the initialization of the locations of facilities, we introduce a physics-inspired strategy for the p-median problem, reaching more stable solutions than the random strategy. The proposed pipeline coupling the classic swap-based method with deep reinforcement learning marks a significant step forward in addressing the practical challenges associated with facility location on graphs.
Reinforcement learning from human feedback (RLHF) is a variant of reinforcement learning (RL) that learns from human feedback instead of relying on an engineered reward function. Building on prior work on the related setting of preference-based reinforcement learning (PbRL), it stands at the intersection of artificial intelligence and human-computer interaction. This positioning offers a promising avenue to enhance the performance and adaptability of intelligent systems while also improving the alignment of their objectives with human values. The training of large language models (LLMs) has impressively demonstrated this potential in recent years, where RLHF played a decisive role in directing the model's capabilities toward human objectives. This article provides a comprehensive overview of the fundamentals of RLHF, exploring the intricate dynamics between RL agents and human input. While recent focus has been on RLHF for LLMs, our survey adopts a broader perspective, examining the diverse applications and wide-ranging impact of the technique. We delve into the core principles that underpin RLHF, shedding light on the symbiotic relationship between algorithms and human feedback, and discuss the main research trends in the field. By synthesizing the current landscape of RLHF research, this article aims to provide researchers as well as practitioners with a comprehensive understanding of this rapidly growing field of research.
A key strategy in societal adaptation to climate change is using alert systems to prompt preventative action and reduce the adverse health impacts of extreme heat events. This paper implements and evaluates reinforcement learning (RL) as a tool to optimize the effectiveness of such systems. Our contributions are threefold. First, we introduce a new publicly available RL environment enabling the evaluation of the effectiveness of heat alert policies to reduce heat-related hospitalizations. The rewards model is trained from a comprehensive dataset of historical weather, Medicare health records, and socioeconomic/geographic features. We use scalable Bayesian techniques tailored to the low-signal effects and spatial heterogeneity present in the data. The transition model uses real historical weather patterns enriched by a data augmentation mechanism based on climate region similarity. Second, we use this environment to evaluate standard RL algorithms in the context of heat alert issuance. Our analysis shows that policy constraints are needed to improve RL's initially poor performance. Third, a post-hoc contrastive analysis provides insight into scenarios where our modified heat alert-RL policies yield significant gains/losses over the current National Weather Service alert policy in the United States.
This work studies the dynamic risk management of the risk-neutral value of the potential credit losses on a portfolio of derivatives. Sensitivities-based hedging of such liability is sub-optimal because of bid-ask costs, pricing models which cannot be completely realistic, and a discontinuity at default time. We leverage recent advances on risk-averse Reinforcement Learning developed specifically for option hedging with an ad hoc practice-aligned objective function aware of pathwise volatility, generalizing them to stochastic horizons. We formalize accurately the evolution of the hedger's portfolio stressing such aspects. We showcase the efficacy of our approach by a numerical study for a portfolio composed of a single FX forward contract.
Recently, model-based reinforcement learning algorithms have demonstrated remarkable efficacy in visual input environments. These approaches begin by constructing a parameterized simulation world model of the real environment through self-supervised learning. By leveraging the imagination of the world model, the agent's policy is enhanced without the constraints of sampling from the real environment. The performance of these algorithms heavily relies on the sequence modeling and generation capabilities of the world model. However, constructing a perfectly accurate model of a complex unknown environment is nearly impossible. Discrepancies between the model and reality may cause the agent to pursue virtual goals, resulting in subpar performance in the real environment. Introducing random noise into model-based reinforcement learning has been proven beneficial. In this work, we introduce Stochastic Transformer-based wORld Model (STORM), an efficient world model architecture that combines the strong sequence modeling and generation capabilities of Transformers with the stochastic nature of variational autoencoders. STORM achieves a mean human performance of $126.7\%$ on the Atari $100$k benchmark, setting a new record among state-of-the-art methods that do not employ lookahead search techniques. Moreover, training an agent with $1.85$ hours of real-time interaction experience on a single NVIDIA GeForce RTX 3090 graphics card requires only $4.3$ hours, showcasing improved efficiency compared to previous methodologies.