Continuous-time Markov decision processes (CTMDPs) are canonical models to express sequential decision-making under dense-time and stochastic environments. When the stochastic evolution of the environment is only available via sampling, model-free reinforcement learning (RL) is the algorithm-of-choice to compute optimal decision sequence. RL, on the other hand, requires the learning objective to be encoded as scalar reward signals. Since doing such translations manually is both tedious and error-prone, a number of techniques have been proposed to translate high-level objectives (expressed in logic or automata formalism) to scalar rewards for discrete-time Markov decision processes (MDPs). Unfortunately, no automatic translation exists for CTMDPs. We consider CTMDP environments against the learning objectives expressed as omega-regular languages. Omega-regular languages generalize regular languages to infinite-horizon specifications and can express properties given in popular linear-time logic LTL. To accommodate the dense-time nature of CTMDPs, we consider two different semantics of omega-regular objectives: 1) satisfaction semantics where the goal of the learner is to maximize the probability of spending positive time in the good states, and 2) expectation semantics where the goal of the learner is to optimize the long-run expected average time spent in the ``good states" of the automaton. We present an approach enabling correct translation to scalar reward signals that can be readily used by off-the-shelf RL algorithms for CTMDPs. We demonstrate the effectiveness of the proposed algorithms by evaluating it on some popular CTMDP benchmarks with omega-regular objectives.
相關內容
We apply reinforcement learning techniques to topic modeling by replacing the variational autoencoder in ProdLDA with a continuous action space reinforcement learning policy. We train the system with a policy gradient algorithm REINFORCE. Additionally, we introduced several modifications: modernize the neural network architecture, weight the ELBO loss, use contextual embeddings, and monitor the learning process via computing topic diversity and coherence for each training step. Experiments are performed on 11 data sets. Our unsupervised model outperforms all other unsupervised models and performs on par with or better than most models using supervised labeling. Our model is outperformed on certain data sets by a model using supervised labeling and contrastive learning. We have also conducted an ablation study to provide empirical evidence of performance improvements from changes we made to ProdLDA and found that the reinforcement learning formulation boosts performance.
Policy optimization methods with function approximation are widely used in multi-agent reinforcement learning. However, it remains elusive how to design such algorithms with statistical guarantees. Leveraging a multi-agent performance difference lemma that characterizes the landscape of multi-agent policy optimization, we find that the localized action value function serves as an ideal descent direction for each local policy. Motivated by the observation, we present a multi-agent PPO algorithm in which the local policy of each agent is updated similarly to vanilla PPO. We prove that with standard regularity conditions on the Markov game and problem-dependent quantities, our algorithm converges to the globally optimal policy at a sublinear rate. We extend our algorithm to the off-policy setting and introduce pessimism to policy evaluation, which aligns with experiments. To our knowledge, this is the first provably convergent multi-agent PPO algorithm in cooperative Markov games.
We tackle the problem of joint frequency and power allocation while emphasizing the generalization capability of a deep reinforcement learning model. Most of the existing methods solve reinforcement learning-based wireless problems for a specific pre-determined wireless network scenario. The performance of a trained agent tends to be very specific to the network and deteriorates when used in a different network operating scenario (e.g., different in size, neighborhood, and mobility, among others). We demonstrate our approach to enhance training to enable a higher generalization capability during inference of the deployed model in a distributed multi-agent setting in a hostile jamming environment. With all these, we show the improved training and inference performance of the proposed methods when tested on previously unseen simulated wireless networks of different sizes and architectures. More importantly, to prove practical impact, the end-to-end solution was implemented on the embedded software-defined radio and validated using over-the-air evaluation.
Most inverse problems from physical sciences are formulated as PDE-constrained optimization problems. This involves identifying unknown parameters in equations by optimizing the model to generate PDE solutions that closely match measured data. The formulation is powerful and widely used in many sciences and engineering fields. However, one crucial assumption is that the unknown parameter must be deterministic. In reality, however, many problems are stochastic in nature, and the unknown parameter is random. The challenge then becomes recovering the full distribution of this unknown random parameter. It is a much more complex task. In this paper, we examine this problem in a general setting. In particular, we conceptualize the PDE solver as a push-forward map that pushes the parameter distribution to the generated data distribution. This way, the SDE-constrained optimization translates to minimizing the distance between the generated distribution and the measurement distribution. We then formulate a gradient-flow equation to seek the ground-truth parameter probability distribution. This opens up a new paradigm for extending many techniques in PDE-constrained optimization to that for systems with stochasticity.
All biological and artificial agents must learn and make decisions given limits on their ability to process information. As such, a general theory of adaptive behavior should be able to account for the complex interactions between an agent's learning history, decisions, and capacity constraints. Recent work in computer science has begun to clarify the principles that shape these dynamics by bridging ideas from reinforcement learning, Bayesian decision-making, and rate-distortion theory. This body of work provides an account of capacity-limited Bayesian reinforcement learning, a unifying normative framework for modeling the effect of processing constraints on learning and action selection. Here, we provide an accessible review of recent algorithms and theoretical results in this setting, paying special attention to how these ideas can be applied to studying questions in the cognitive and behavioral sciences.
While off-policy reinforcement learning (RL) algorithms are sample efficient due to gradient-based updates and data reuse in the replay buffer, they struggle with convergence to local optima due to limited exploration. On the other hand, population-based algorithms offer a natural exploration strategy, but their heuristic black-box operators are inefficient. Recent algorithms have integrated these two methods, connecting them through a shared replay buffer. However, the effect of using diverse data from population optimization iterations on off-policy RL algorithms has not been thoroughly investigated. In this paper, we first analyze the use of off-policy RL algorithms in combination with population-based algorithms, showing that the use of population data could introduce an overlooked error and harm performance. To test this, we propose a uniform and scalable training design and conduct experiments on our tailored framework in robot locomotion tasks from the OpenAI gym. Our results substantiate that using population data in off-policy RL can cause instability during training and even degrade performance. To remedy this issue, we further propose a double replay buffer design that provides more on-policy data and show its effectiveness through experiments. Our results offer practical insights for training these hybrid methods.
Controller synthesis is in essence a case of model-based planning for non-deterministic environments in which plans (actually ''strategies'') are meant to preserve system goals indefinitely. In the case of supervisory control environments are specified as the parallel composition of state machines and valid strategies are required to be ''non-blocking'' (i.e., always enabling the environment to reach certain marked states) in addition to safe (i.e., keep the system within a safe zone). Recently, On-the-fly Directed Controller Synthesis techniques were proposed to avoid the exploration of the entire -and exponentially large-environment space, at the cost of non-maximal permissiveness, to either find a strategy or conclude that there is none. The incremental exploration of the plant is currently guided by a domain-independent human-designed heuristic. In this work, we propose a new method for obtaining heuristics based on Reinforcement Learning (RL). The synthesis algorithm is thus framed as an RL task with an unbounded action space and a modified version of DQN is used. With a simple and general set of features that abstracts both states and actions, we show that it is possible to learn heuristics on small versions of a problem that generalize to the larger instances, effectively doing zero-shot policy transfer. Our agents learn from scratch in a highly partially observable RL task and outperform the existing heuristic overall, in instances unseen during training.
While Reinforcement Learning (RL) achieves tremendous success in sequential decision-making problems of many domains, it still faces key challenges of data inefficiency and the lack of interpretability. Interestingly, many researchers have leveraged insights from the causality literature recently, bringing forth flourishing works to unify the merits of causality and address well the challenges from RL. As such, it is of great necessity and significance to collate these Causal Reinforcement Learning (CRL) works, offer a review of CRL methods, and investigate the potential functionality from causality toward RL. In particular, we divide existing CRL approaches into two categories according to whether their causality-based information is given in advance or not. We further analyze each category in terms of the formalization of different models, ranging from the Markov Decision Process (MDP), Partially Observed Markov Decision Process (POMDP), Multi-Arm Bandits (MAB), and Dynamic Treatment Regime (DTR). Moreover, we summarize the evaluation matrices and open sources while we discuss emerging applications, along with promising prospects for the future development of CRL.
We introduce DeepNash, an autonomous agent capable of learning to play the imperfect information game Stratego from scratch, up to a human expert level. Stratego is one of the few iconic board games that Artificial Intelligence (AI) has not yet mastered. This popular game has an enormous game tree on the order of $10^{535}$ nodes, i.e., $10^{175}$ times larger than that of Go. It has the additional complexity of requiring decision-making under imperfect information, similar to Texas hold'em poker, which has a significantly smaller game tree (on the order of $10^{164}$ nodes). Decisions in Stratego are made over a large number of discrete actions with no obvious link between action and outcome. Episodes are long, with often hundreds of moves before a player wins, and situations in Stratego can not easily be broken down into manageably-sized sub-problems as in poker. For these reasons, Stratego has been a grand challenge for the field of AI for decades, and existing AI methods barely reach an amateur level of play. DeepNash uses a game-theoretic, model-free deep reinforcement learning method, without search, that learns to master Stratego via self-play. The Regularised Nash Dynamics (R-NaD) algorithm, a key component of DeepNash, converges to an approximate Nash equilibrium, instead of 'cycling' around it, by directly modifying the underlying multi-agent learning dynamics. DeepNash beats existing state-of-the-art AI methods in Stratego and achieved a yearly (2022) and all-time top-3 rank on the Gravon games platform, competing with human expert players.
We propose a new method for event extraction (EE) task based on an imitation learning framework, specifically, inverse reinforcement learning (IRL) via generative adversarial network (GAN). The GAN estimates proper rewards according to the difference between the actions committed by the expert (or ground truth) and the agent among complicated states in the environment. EE task benefits from these dynamic rewards because instances and labels yield to various extents of difficulty and the gains are expected to be diverse -- e.g., an ambiguous but correctly detected trigger or argument should receive high gains -- while the traditional RL models usually neglect such differences and pay equal attention on all instances. Moreover, our experiments also demonstrate that the proposed framework outperforms state-of-the-art methods, without explicit feature engineering.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191