Reinforcement learning (RL) trains an agent from experiences interacting with the environment. In scenarios where online interactions are impractical, offline RL, which trains the agent using pre-collected datasets, has become popular. While this new paradigm presents remarkable effectiveness across various real-world domains, like healthcare and energy management, there is a growing demand to enable agents to rapidly and completely eliminate the influence of specific trajectories from both the training dataset and the trained agents. To meet this problem, this paper advocates Trajdeleter, the first practical approach to trajectory unlearning for offline RL agents. The key idea of Trajdeleter is to guide the agent to demonstrate deteriorating performance when it encounters states associated with unlearning trajectories. Simultaneously, it ensures the agent maintains its original performance level when facing other remaining trajectories. Additionally, we introduce Trajauditor, a simple yet efficient method to evaluate whether Trajdeleter successfully eliminates the specific trajectories of influence from the offline RL agent. Extensive experiments conducted on six offline RL algorithms and three tasks demonstrate that Trajdeleter requires only about 1.5% of the time needed for retraining from scratch. It effectively unlearns an average of 94.8% of the targeted trajectories yet still performs well in actual environment interactions after unlearning. The replication package and agent parameters are available online.
This paper studies continuous-time risk-sensitive reinforcement learning (RL) under the entropy-regularized, exploratory diffusion process formulation with the exponential-form objective. The risk-sensitive objective arises either as the agent's risk attitude or as a distributionally robust approach against the model uncertainty. Owing to the martingale perspective in Jia and Zhou (2023) the risk-sensitive RL problem is shown to be equivalent to ensuring the martingale property of a process involving both the value function and the q-function, augmented by an additional penalty term: the quadratic variation of the value process, capturing the variability of the value-to-go along the trajectory. This characterization allows for the straightforward adaptation of existing RL algorithms developed for non-risk-sensitive scenarios to incorporate risk sensitivity by adding the realized variance of the value process. Additionally, I highlight that the conventional policy gradient representation is inadequate for risk-sensitive problems due to the nonlinear nature of quadratic variation; however, q-learning offers a solution and extends to infinite horizon settings. Finally, I prove the convergence of the proposed algorithm for Merton's investment problem and quantify the impact of temperature parameter on the behavior of the learning procedure. I also conduct simulation experiments to demonstrate how risk-sensitive RL improves the finite-sample performance in the linear-quadratic control problem.
In this work, we propose a new setting of continual learning: data-incremental continual offline reinforcement learning (DICORL), in which an agent is asked to learn a sequence of datasets of a single offline reinforcement learning (RL) task continually, instead of learning a sequence of offline RL tasks with respective datasets. Then, we propose that this new setting will introduce a unique challenge to continual learning: active forgetting, which means that the agent will forget the learnt skill actively. The main reason for active forgetting is conservative learning used by offline RL, which is used to solve the overestimation problem. With conservative learning, the offline RL method will suppress the value of all actions, learnt or not, without selection, unless it is in the just learning dataset. Therefore, inferior data may overlay premium data because of the learning sequence. To solve this problem, we propose a new algorithm, called experience-replay-based ensemble implicit Q-learning (EREIQL), which introduces multiple value networks to reduce the initial value and avoid using conservative learning, and the experience replay to relieve catastrophic forgetting. Our experiments show that EREIQL relieves active forgetting in DICORL and performs well.
Recent advancements in LLMs have shown their significant potential in tasks like text summarization and generation. Yet, they often encounter difficulty while solving complex physics problems that require arithmetic calculation and a good understanding of concepts. Moreover, many physics problems include images that contain important details required to understand the problem's context. We propose an LMM-based chatbot to answer multimodal physics MCQs. For domain adaptation, we utilize the MM-PhyQA dataset comprising Indian high school-level multimodal physics problems. To improve the LMM's performance, we experiment with two techniques, RLHF (Reinforcement Learning from Human Feedback) and Image Captioning. In image captioning, we add a detailed explanation of the diagram in each image, minimizing hallucinations and image processing errors. We further explore the integration of Reinforcement Learning from Human Feedback (RLHF) methodology inspired by the ranking approach in RLHF to enhance the human-like problem-solving abilities of the models. The RLHF approach incorporates human feedback into the learning process of LLMs, improving the model's problem-solving skills, truthfulness, and reasoning capabilities, minimizing the hallucinations in the answers, and improving the quality instead of using vanilla-supervised fine-tuned models. We employ the LLaVA open-source model to answer multimodal physics MCQs and compare the performance with and without using RLHF.
Visual Reinforcement Learning is a popular and powerful framework that takes full advantage of the Deep Learning breakthrough. However, it is also known that variations in the input (e.g., different colors of the panorama due to the season of the year) or the task (e.g., changing the speed limit for a car to respect) could require complete retraining of the agents. In this work, we leverage recent developments in unifying latent representations to demonstrate that it is possible to combine the components of an agent, rather than retrain it from scratch. We build upon the recent relative representations framework and adapt it for Visual RL. This allows us to create completely new agents capable of handling environment-task combinations never seen during training. Our work paves the road toward a more accessible and flexible use of reinforcement learning.
The widespread use of knowledge graphs in various fields has brought about a challenge in effectively integrating and updating information within them. When it comes to incorporating contexts, conventional methods often rely on rules or basic machine learning models, which may not fully grasp the complexity and fluidity of context information. This research suggests an approach based on reinforcement learning (RL), specifically utilizing Deep Q Networks (DQN) to enhance the process of integrating contexts into knowledge graphs. By considering the state of the knowledge graph as environment states defining actions as operations for integrating contexts and using a reward function to gauge the improvement in knowledge graph quality post-integration, this method aims to automatically develop strategies for optimal context integration. Our DQN model utilizes networks as function approximators, continually updating Q values to estimate the action value function, thus enabling effective integration of intricate and dynamic context information. Initial experimental findings show that our RL method outperforms techniques in achieving precise context integration across various standard knowledge graph datasets, highlighting the potential and effectiveness of reinforcement learning in enhancing and managing knowledge graphs.
Policy gradient methods in actor-critic reinforcement learning (RL) have become perhaps the most promising approaches to solving continuous optimal control problems. However, the trial-and-error nature of RL and the inherent randomness associated with solution approximations cause variations in the learned optimal values and policies. This has significantly hindered their successful deployment in real life applications where control responses need to meet dynamic performance criteria deterministically. Here we propose a novel phased actor in actor-critic (PAAC) method, aiming at improving policy gradient estimation and thus the quality of the control policy. Specifically, PAAC accounts for both $Q$ value and TD error in its actor update. We prove qualitative properties of PAAC for learning convergence of the value and policy, solution optimality, and stability of system dynamics. Additionally, we show variance reduction in policy gradient estimation. PAAC performance is systematically and quantitatively evaluated in this study using DeepMind Control Suite (DMC). Results show that PAAC leads to significant performance improvement measured by total cost, learning variance, robustness, learning speed and success rate. As PAAC can be piggybacked onto general policy gradient learning frameworks, we select well-known methods such as direct heuristic dynamic programming (dHDP), deep deterministic policy gradient (DDPG) and their variants to demonstrate the effectiveness of PAAC. Consequently we provide a unified view on these related policy gradient algorithms.
Reinforcement learning has recently gained unprecedented popularity, yet it still grapples with sample inefficiency. Addressing this challenge, federated reinforcement learning (FedRL) has emerged, wherein agents collaboratively learn a single policy by aggregating local estimations. However, this aggregation step incurs significant communication costs. In this paper, we propose CompFedRL, a communication-efficient FedRL approach incorporating both \textit{periodic aggregation} and (direct/error-feedback) compression mechanisms. Specifically, we consider compressed federated $Q$-learning with a generative model setup, where a central server learns an optimal $Q$-function by periodically aggregating compressed $Q$-estimates from local agents. For the first time, we characterize the impact of these two mechanisms (which have remained elusive) by providing a finite-time analysis of our algorithm, demonstrating strong convergence behaviors when utilizing either direct or error-feedback compression. Our bounds indicate improved solution accuracy concerning the number of agents and other federated hyperparameters while simultaneously reducing communication costs. To corroborate our theory, we also conduct in-depth numerical experiments to verify our findings, considering Top-$K$ and Sparsified-$K$ sparsification operators.
Cooperative Multi-Agent Reinforcement Learning (MARL) necessitates seamless collaboration among agents, often represented by an underlying relation graph. Existing methods for learning this graph primarily focus on agent-pair relations, neglecting higher-order relationships. While several approaches attempt to extend cooperation modelling to encompass behaviour similarities within groups, they commonly fall short in concurrently learning the latent graph, thereby constraining the information exchange among partially observed agents. To overcome these limitations, we present a novel approach to infer the Group-Aware Coordination Graph (GACG), which is designed to capture both the cooperation between agent pairs based on current observations and group-level dependencies from behaviour patterns observed across trajectories. This graph is further used in graph convolution for information exchange between agents during decision-making. To further ensure behavioural consistency among agents within the same group, we introduce a group distance loss, which promotes group cohesion and encourages specialization between groups. Our evaluations, conducted on StarCraft II micromanagement tasks, demonstrate GACG's superior performance. An ablation study further provides experimental evidence of the effectiveness of each component of our method.
In sequential decision-making problems involving sensitive attributes like race and gender, reinforcement learning (RL) agents must carefully consider long-term fairness while maximizing returns. Recent works have proposed many different types of fairness notions, but how unfairness arises in RL problems remains unclear. In this paper, we address this gap in the literature by investigating the sources of inequality through a causal lens. We first analyse the causal relationships governing the data generation process and decompose the effect of sensitive attributes on long-term well-being into distinct components. We then introduce a novel notion called dynamics fairness, which explicitly captures the inequality stemming from environmental dynamics, distinguishing it from those induced by decision-making or inherited from the past. This notion requires evaluating the expected changes in the next state and the reward induced by changing the value of the sensitive attribute while holding everything else constant. To quantitatively evaluate this counterfactual concept, we derive identification formulas that allow us to obtain reliable estimations from data. Extensive experiments demonstrate the effectiveness of the proposed techniques in explaining, detecting, and reducing inequality in reinforcement learning. We publicly release code at https://github.com/familyld/InsightFair.
Embracing the pursuit of intrinsically explainable reinforcement learning raises crucial questions: what distinguishes explainability from interpretability? Should explainable and interpretable agents be developed outside of domains where transparency is imperative? What advantages do interpretable policies offer over neural networks? How can we rigorously define and measure interpretability in policies, without user studies? What reinforcement learning paradigms,are the most suited to develop interpretable agents? Can Markov Decision Processes integrate interpretable state representations? In addition to motivate an Interpretable RL community centered around the aforementioned questions, we propose the first venue dedicated to Interpretable RL: the InterpPol Workshop.
The rapid growth of machine learning (ML) has led to an increased demand for computational power, resulting in larger data centers (DCs) and higher energy consumption. To address this issue and reduce carbon emissions, intelligent design and control of DC components such as IT servers, cabinets, HVAC cooling, flexible load shifting, and battery energy storage are essential. However, the complexity of designing and controlling them in tandem presents a significant challenge. While some individual components like CFD-based design and Reinforcement Learning (RL) based HVAC control have been researched, there's a gap in the holistic design and optimization covering all elements simultaneously. To tackle this, we've developed DCRL-Green, a multi-agent RL environment that empowers the ML community to design data centers and research, develop, and refine RL controllers for carbon footprint reduction in DCs. It is a flexible, modular, scalable, and configurable platform that can handle large High Performance Computing (HPC) clusters. Furthermore, in its default setup, DCRL-Green provides a benchmark for evaluating single as well as multi-agent RL algorithms. It easily allows users to subclass the default implementations and design their own control approaches, encouraging community development for sustainable data centers. Open Source Link: https://github.com/HewlettPackard/dc-rl
Linear Temporal Logic (LTL) offers a precise means for constraining the behavior of reinforcement learning agents. However, in many settings where both satisfaction and optimality conditions are present, LTL is insufficient to capture both. Instead, LTL-constrained policy optimization, where the goal is to optimize a scalar reward under LTL constraints, is needed. This constrained optimization problem proves difficult in deep Reinforcement Learning (DRL) settings, where learned policies often ignore the LTL constraint due to the sparse nature of LTL satisfaction. To alleviate the sparsity issue, we introduce Cycle Experience Replay (CyclER), a novel reward shaping technique that exploits the underlying structure of the LTL constraint to guide a policy towards satisfaction by encouraging partial behaviors compliant with the constraint. We provide a theoretical guarantee that optimizing CyclER will achieve policies that satisfy the LTL constraint with near-optimal probability. We evaluate CyclER in three continuous control domains. Our experimental results show that optimizing CyclER in tandem with the existing scalar reward outperforms existing reward-shaping methods at finding performant LTL-satisfying policies.
Text-based reinforcement learning involves an agent interacting with a fictional environment using observed text and admissible actions in natural language to complete a task. Previous works have shown that agents can succeed in text-based interactive environments even in the complete absence of semantic understanding or other linguistic capabilities. The success of these agents in playing such games suggests that semantic understanding may not be important for the task. This raises an important question about the benefits of LMs in guiding the agents through the game states. In this work, we show that rich semantic understanding leads to efficient training of text-based RL agents. Moreover, we describe the occurrence of semantic degeneration as a consequence of inappropriate fine-tuning of language models in text-based reinforcement learning (TBRL). Specifically, we describe the shift in the semantic representation of words in the LM, as well as how it affects the performance of the agent in tasks that are semantically similar to the training games. We believe these results may help develop better strategies to fine-tune agents in text-based RL scenarios.
We present the first study on provably efficient randomized exploration in cooperative multi-agent reinforcement learning (MARL). We propose a unified algorithm framework for randomized exploration in parallel Markov Decision Processes (MDPs), and two Thompson Sampling (TS)-type algorithms, CoopTS-PHE and CoopTS-LMC, incorporating the perturbed-history exploration (PHE) strategy and the Langevin Monte Carlo exploration (LMC) strategy respectively, which are flexible in design and easy to implement in practice. For a special class of parallel MDPs where the transition is (approximately) linear, we theoretically prove that both CoopTS-PHE and CoopTS-LMC achieve a $\widetilde{\mathcal{O}}(d^{3/2}H^2\sqrt{MK})$ regret bound with communication complexity $\widetilde{\mathcal{O}}(dHM^2)$, where $d$ is the feature dimension, $H$ is the horizon length, $M$ is the number of agents, and $K$ is the number of episodes. This is the first theoretical result for randomized exploration in cooperative MARL. We evaluate our proposed method on multiple parallel RL environments, including a deep exploration problem (\textit{i.e.,} $N$-chain), a video game, and a real-world problem in energy systems. Our experimental results support that our framework can achieve better performance, even under conditions of misspecified transition models. Additionally, we establish a connection between our unified framework and the practical application of federated learning.
During their operation, due to shifts in environmental conditions, devices undergo various forms of detuning from their optimal settings. Typically, this is addressed through control loops, which monitor variables and the device performance, to maintain settings at their optimal values. Quantum devices are particularly challenging since their functionality relies on precisely tuning their parameters. At the same time, the detailed modeling of the environmental behavior is often computationally unaffordable, while a direct measure of the parameters defining the system state is costly and introduces extra noise in the mechanism. In this study, we investigate the application of reinforcement learning techniques to develop a model-free control loop for continuous recalibration of quantum device parameters. Furthermore, we explore the advantages of incorporating minimal environmental noise models. As an example, the application to numerical simulations of a Kennedy receiver-based long-distance quantum communication protocol is presented.
Reinforcement learning is a powerful framework aiming to determine optimal behavior in highly complex decision-making scenarios. This objective can be achieved using policy iteration, which requires to solve a typically large linear system of equations. We propose the variational quantum policy iteration (VarQPI) algorithm, realizing this step with a NISQ-compatible quantum-enhanced subroutine. Its scalability is supported by an analysis of the structure of generic reinforcement learning environments, laying the foundation for potential quantum advantage with utility-scale quantum computers. Furthermore, we introduce the warm-start initialization variant (WS-VarQPI) that significantly reduces resource overhead. The algorithm solves a large FrozenLake environment with an underlying 256x256-dimensional linear system, indicating its practical robustness.
Offline reinforcement learning (RL) aims to learn policies from static datasets of previously collected trajectories. Existing methods for offline RL either constrain the learned policy to the support of offline data or utilize model-based virtual environments to generate simulated rollouts. However, these methods suffer from (i) poor generalization to unseen states; and (ii) trivial improvement from low-qualified rollout simulation. In this paper, we propose offline trajectory generalization through world transformers for offline reinforcement learning (OTTO). Specifically, we use casual Transformers, a.k.a. World Transformers, to predict state dynamics and the immediate reward. Then we propose four strategies to use World Transformers to generate high-rewarded trajectory simulation by perturbing the offline data. Finally, we jointly use offline data with simulated data to train an offline RL algorithm. OTTO serves as a plug-in module and can be integrated with existing offline RL methods to enhance them with better generalization capability of transformers and high-rewarded data augmentation. Conducting extensive experiments on D4RL benchmark datasets, we verify that OTTO significantly outperforms state-of-the-art offline RL methods.
From a visual perception perspective, modern graphical user interfaces (GUIs) comprise a complex graphics-rich two-dimensional visuospatial arrangement of text, images, and interactive objects such as buttons and menus. While existing models can accurately predict regions and objects that are likely to attract attention ``on average'', so far there is no scanpath model capable of predicting scanpaths for an individual. To close this gap, we introduce EyeFormer, which leverages a Transformer architecture as a policy network to guide a deep reinforcement learning algorithm that controls gaze locations. Our model has the unique capability of producing personalized predictions when given a few user scanpath samples. It can predict full scanpath information, including fixation positions and duration, across individuals and various stimulus types. Additionally, we demonstrate applications in GUI layout optimization driven by our model. Our software and models will be publicly available.
This paper presents the first algorithm for model-based offline quantum reinforcement learning and demonstrates its functionality on the cart-pole benchmark. The model and the policy to be optimized are each implemented as variational quantum circuits. The model is trained by gradient descent to fit a pre-recorded data set. The policy is optimized with a gradient-free optimization scheme using the return estimate given by the model as the fitness function. This model-based approach allows, in principle, full realization on a quantum computer during the optimization phase and gives hope that a quantum advantage can be achieved as soon as sufficiently powerful quantum computers are available.